snn_proto_exo.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "toc_visible": true,
      "mount_file_id": "1ApkctSJPWFx1KZuB0M_-N7GeajLuFz9X",
      "authorship_tag": "ABX9TyNjz0UMuqFfPlLxiUBJD/hE",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/peppermintbird/nabs-prototype/blob/main/snn_proto_exo.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Note and objective"
      ],
      "metadata": {
        "id": "D9ysFZoWIBNI"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "This notebook is a compendium of snnTorch's tutorials, Kaggle's competition datasets, and my own coding, notes, and experiments.  "
      ],
      "metadata": {
        "id": "Lh3lVSi1c6aI"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "However, the main structure of this model is from [snnTorch's dataset tutorial series: Exoplanet Hunter](https://snntorch.readthedocs.io/en/latest/tutorials/tutorial_exoplanet_hunter.html), written by Ruhai Lin, Aled dela Cruz, and Karina Aguilar."
      ],
      "metadata": {
        "id": "w9YnK1q5IT5W"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Here, I aim to provide a simple overview of how SNN can serve as an architecture for an exoplanet detection deep learning model. I'll then explain how the code relates to the data analysis.\n",
        "\n"
      ],
      "metadata": {
        "id": "FiX_1XS4Fop3"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "---"
      ],
      "metadata": {
        "id": "6vsREDL1-SfU"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Data source"
      ],
      "metadata": {
        "id": "H_nWUpb7-VhH"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "In the context of promising space missions like NASA's Kepler and TESS missions, there were three main types of astronomical observation data—target pixel files, light curves, and tabular stellar information.\n"
      ],
      "metadata": {
        "id": "f14sY7lm-a72"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "In this small model, because of time and resources constraints, we will use light curve data from the Kepler space telescope that was cleaned and organized in this [Kaggle](https://www.kaggle.com/datasets/keplersmachines/kepler-labelled-time-series-data) dataset."
      ],
      "metadata": {
        "id": "0D3l7nXI_mBc"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Kepler was a space telescope mission launched by NASA with the specially designed to discover Earth-sized exoplanets orbiting other stars in the Milky Way within their habitable zones.\n"
      ],
      "metadata": {
        "id": "blxm5g6jCJnV"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "It operated from 2009 to 2018. Kepler used the transit method to detect exoplanets, providing a dataset of photometric time series data for numerous stars. This data, which has already been denoised and cleaned of artifacts, is publicly available in the [Mikulski Archive](https://archive.stsci.edu/missions-and-data/kepler) and in the [NASA Exoplanet Archive](https://exoplanetarchive.ipac.caltech.edu/docs/KeplerMission.html)."
      ],
      "metadata": {
        "id": "MVGM6GBPtPFq"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "> **Acknowledging limitations**\n",
        "\n",
        "- Kepler primarily provides photometric measurements (brightness) rather than detailed spectral information.\n",
        "- Since Kepler has a **large** dataset, preprocessing and cleaning the data is a challenge within this project’s constraints. In this model, I’m using WΔ's cleaned dataset, which is hosted on [Kaggle](https://www.kaggle.com/datasets/keplersmachines/kepler-labelled-time-series-data/data). There is also more information on WΔ’s [GitHub repository](https://github.com/winterdelta/KeplerAI).\n",
        "- According to WΔ, Kepler's Campaign 3 data is utilized to focus on a subset of data considered to be of higher reliability. Even so, it is unclear what selection criteria or processing steps WΔ applies to the dataset curation.\n"
      ],
      "metadata": {
        "id": "5YJvuvD1EiwM"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "---"
      ],
      "metadata": {
        "id": "RTJN9WrT-jL6"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Data setup and preprocessing"
      ],
      "metadata": {
        "id": "9W_ySf8jDiCY"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "First of all, ensure you have the needed requirements to use snnTorch. [By following PyPi's latest version of snnTorch](https://pypi.org/project/snntorch/), you should find a quick start in how to setup the package."
      ],
      "metadata": {
        "id": "hMF-6lpqZUuQ"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "* Install or collect the following packages and libraries; ignore those you already have installed. Optionally, add the `--quiet` flag to avoid cumbersome output messages."
      ],
      "metadata": {
        "id": "StEYromq4qDJ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!pip install snntorch\n",
        "!pip install torch\n",
        "!pip install imblearn\n",
        "!pip install torchvision\n",
        "!pip install matplotlib\n",
        "!pip install numpy\n",
        "!pip install pandas\n",
        "!pip install seaborn\n",
        "!pip install plotly.express\n",
        "\n"
      ],
      "metadata": {
        "id": "8TWsOdSE3j1f"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "* Now import those packages."
      ],
      "metadata": {
        "id": "WtutQIg85ViT"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# importing snnTorch\n",
        "import snntorch as snn\n",
        "from snntorch import surrogate\n",
        "\n",
        "# setting pyTorch\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim\n",
        "import torch.nn.functional as F\n",
        "from torch.utils.data import DataLoader, Dataset\n",
        "from torchvision import transforms\n",
        "\n",
        "# SMOTE (oversampling to balance the dataset)\n",
        "from imblearn.over_sampling import SMOTE\n",
        "from collections import Counter\n",
        "\n",
        "# plot\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import seaborn as sns\n",
        "import plotly.express as px\n",
        "import plotly.graph_objects as go\n",
        "from plotly.subplots import make_subplots\n",
        "\n",
        "# metric (AUC, ROC, sensitivity & specificity)\n",
        "from sklearn.metrics import confusion_matrix\n",
        "from sklearn.metrics import roc_auc_score\n"
      ],
      "metadata": {
        "id": "wPKjKEKAyN0t"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "* Download the dataset from [Kaggle](https://www.kaggle.com/datasets/keplersmachines/kepler-labelled-time-series-data/data) and then export it to Google Drive if you're using Google Colab. To ensure your files are accessible on Google Colab, mount them at `/content/drive` to connect your Drive to Google Colab.\n",
        "\n",
        "  * For this small project we will use tested datasets:\n",
        "    1. `exoTest.csv`\n",
        "    2. `exoTrain.csv`"
      ],
      "metadata": {
        "id": "eCUjdcZD5_4T"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "VK2uXgD_wwXl",
        "outputId": "84a21b11-316b-47e4-aeef-6c58a200f339"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
          ]
        }
      ],
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "%cd '/content/drive/MyDrive/SNN'\n",
        "\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CMv3ZzpR-MuZ",
        "outputId": "410bd186-23a8-49df-fcd6-0a0d20b7c20f"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "/content/drive/MyDrive/SNN\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Just to be _sure_, let's see if the files can indeed be found."
      ],
      "metadata": {
        "id": "bzBUP_iqCnsb"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!ls exoTest.csv exoTrain.csv"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gE9MpoFLCdza",
        "outputId": "594aab9a-5d10-4fe4-d111-edc6e184af64"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "exoTest.csv  exoTrain.csv\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "- The next step is optional, but I think it's good if you didn't get to understand the data you're handling."
      ],
      "metadata": {
        "id": "ulVjrzCVERYv"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "> **Reminder:**\n",
        "\n",
        "Var 1 = Contain the label attribute stating:\n",
        "      \n",
        "  - 1 = not exoplanet\n",
        "\n",
        "  - 2 = exoplanet\n",
        "\n",
        " If 1 and 2 are somewhat inconvenient, don't worry, we will change these label values into more intuitive ones, like 0 and 1.\n",
        "\n",
        "\n",
        "Var 2 - 3098 = Contain the flux values measured at different points in time."
      ],
      "metadata": {
        "id": "jw22l2nZGh4S"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "* Read your datasets by stating:   "
      ],
      "metadata": {
        "id": "LNEKWG47E246"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "exoTest = pd.read_csv('/content/drive/MyDrive/SNN/exoTest.csv')\n",
        "exoTrain = pd.read_csv('/content/drive/MyDrive/SNN/exoTrain.csv')\n",
        "\n",
        "exoTest.tail(10) # you can input here the number of rows you wish to peruse"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 412
        },
        "id": "Ey5TdBKHFYWQ",
        "outputId": "f200b757-e251-4e80-c5f3-a8f1860a043e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "     LABEL   FLUX.1   FLUX.2   FLUX.3   FLUX.4   FLUX.5   FLUX.6   FLUX.7  \\\n",
              "560      1     5.66     7.83    -4.39    -0.62    -5.23     0.34   -13.64   \n",
              "561      1     4.36     2.22     5.61    -1.44    -0.05    -0.22     1.28   \n",
              "562      1   -79.95   -79.13   -81.07   -72.25   -79.71   -70.60   -90.33   \n",
              "563      1   -63.42   -59.78   -64.16   -61.07   -41.74   -58.99   -35.94   \n",
              "564      1   144.47    20.09   147.47   109.18   100.87    24.87   115.03   \n",
              "565      1   374.46   326.06   319.87   338.23   251.54   209.84   186.35   \n",
              "566      1    -0.36     4.96     6.25     4.20     8.26    -9.53   -10.10   \n",
              "567      1   -54.01   -44.13   -41.23   -42.82   -39.47   -24.88   -31.14   \n",
              "568      1    91.36    85.60    48.81    48.69    70.05    22.30    11.63   \n",
              "569      1  3071.19  2782.53  2608.69  2325.47  2089.37  1769.56  1421.09   \n",
              "\n",
              "      FLUX.8  FLUX.9  ...  FLUX.3188  FLUX.3189  FLUX.3190  FLUX.3191  \\\n",
              "560   -11.26   -9.96  ...      15.15      24.89      23.41      26.33   \n",
              "561   -13.20   -5.83  ...       2.94      -2.90       2.12      11.26   \n",
              "562   -88.66  -68.15  ...       5.40      19.51      -0.15       1.67   \n",
              "563   -43.15  -24.87  ...     -13.98       0.62      -5.53     -14.26   \n",
              "564   106.37   48.87  ...     -57.53     -41.94    -165.94    -265.13   \n",
              "565   167.46  135.45  ...    -123.55    -166.90    -222.44    -209.71   \n",
              "566    -4.54  -11.55  ...     -12.40      -5.99     -17.94     -11.96   \n",
              "567   -24.71  -13.12  ...      -0.73      -1.64       1.58      -4.82   \n",
              "568    37.86   28.27  ...       2.44      11.53     -16.42     -17.86   \n",
              "569  1142.09  902.31  ...     695.41     865.97     882.41    1203.06   \n",
              "\n",
              "     FLUX.3192  FLUX.3193  FLUX.3194  FLUX.3195  FLUX.3196  FLUX.3197  \n",
              "560      44.37      46.15      37.28      45.60      41.66      39.35  \n",
              "561      12.81       6.03      -6.13     -10.46      -5.36     -14.23  \n",
              "562       3.20       7.25      -8.25      -7.91       4.45       6.58  \n",
              "563     -23.69     -37.84      -2.37       6.62      12.02       2.33  \n",
              "564    -263.35    -376.53    -176.22    -152.07    -216.60    -202.97  \n",
              "565    -180.16    -166.83    -235.66    -213.63    -205.99    -194.07  \n",
              "566     -12.11     -13.68      -3.59      -5.32     -10.98     -11.24  \n",
              "567     -11.93     -17.14      -4.25       5.47      14.46      18.70  \n",
              "568      21.10     -10.25     -37.06      -8.43      -6.48      17.60  \n",
              "569    1293.03    1354.41    -192.81    -277.22     -69.63     121.56  \n",
              "\n",
              "[10 rows x 3198 columns]"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-19ed28f1-fca9-4a0f-9dc7-28668b4e7275\" class=\"colab-df-container\">\n",
              "    <div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>LABEL</th>\n",
              "      <th>FLUX.1</th>\n",
              "      <th>FLUX.2</th>\n",
              "      <th>FLUX.3</th>\n",
              "      <th>FLUX.4</th>\n",
              "      <th>FLUX.5</th>\n",
              "      <th>FLUX.6</th>\n",
              "      <th>FLUX.7</th>\n",
              "      <th>FLUX.8</th>\n",
              "      <th>FLUX.9</th>\n",
              "      <th>...</th>\n",
              "      <th>FLUX.3188</th>\n",
              "      <th>FLUX.3189</th>\n",
              "      <th>FLUX.3190</th>\n",
              "      <th>FLUX.3191</th>\n",
              "      <th>FLUX.3192</th>\n",
              "      <th>FLUX.3193</th>\n",
              "      <th>FLUX.3194</th>\n",
              "      <th>FLUX.3195</th>\n",
              "      <th>FLUX.3196</th>\n",
              "      <th>FLUX.3197</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>560</th>\n",
              "      <td>1</td>\n",
              "      <td>5.66</td>\n",
              "      <td>7.83</td>\n",
              "      <td>-4.39</td>\n",
              "      <td>-0.62</td>\n",
              "      <td>-5.23</td>\n",
              "      <td>0.34</td>\n",
              "      <td>-13.64</td>\n",
              "      <td>-11.26</td>\n",
              "      <td>-9.96</td>\n",
              "      <td>...</td>\n",
              "      <td>15.15</td>\n",
              "      <td>24.89</td>\n",
              "      <td>23.41</td>\n",
              "      <td>26.33</td>\n",
              "      <td>44.37</td>\n",
              "      <td>46.15</td>\n",
              "      <td>37.28</td>\n",
              "      <td>45.60</td>\n",
              "      <td>41.66</td>\n",
              "      <td>39.35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>561</th>\n",
              "      <td>1</td>\n",
              "      <td>4.36</td>\n",
              "      <td>2.22</td>\n",
              "      <td>5.61</td>\n",
              "      <td>-1.44</td>\n",
              "      <td>-0.05</td>\n",
              "      <td>-0.22</td>\n",
              "      <td>1.28</td>\n",
              "      <td>-13.20</td>\n",
              "      <td>-5.83</td>\n",
              "      <td>...</td>\n",
              "      <td>2.94</td>\n",
              "      <td>-2.90</td>\n",
              "      <td>2.12</td>\n",
              "      <td>11.26</td>\n",
              "      <td>12.81</td>\n",
              "      <td>6.03</td>\n",
              "      <td>-6.13</td>\n",
              "      <td>-10.46</td>\n",
              "      <td>-5.36</td>\n",
              "      <td>-14.23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>562</th>\n",
              "      <td>1</td>\n",
              "      <td>-79.95</td>\n",
              "      <td>-79.13</td>\n",
              "      <td>-81.07</td>\n",
              "      <td>-72.25</td>\n",
              "      <td>-79.71</td>\n",
              "      <td>-70.60</td>\n",
              "      <td>-90.33</td>\n",
              "      <td>-88.66</td>\n",
              "      <td>-68.15</td>\n",
              "      <td>...</td>\n",
              "      <td>5.40</td>\n",
              "      <td>19.51</td>\n",
              "      <td>-0.15</td>\n",
              "      <td>1.67</td>\n",
              "      <td>3.20</td>\n",
              "      <td>7.25</td>\n",
              "      <td>-8.25</td>\n",
              "      <td>-7.91</td>\n",
              "      <td>4.45</td>\n",
              "      <td>6.58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>563</th>\n",
              "      <td>1</td>\n",
              "      <td>-63.42</td>\n",
              "      <td>-59.78</td>\n",
              "      <td>-64.16</td>\n",
              "      <td>-61.07</td>\n",
              "      <td>-41.74</td>\n",
              "      <td>-58.99</td>\n",
              "      <td>-35.94</td>\n",
              "      <td>-43.15</td>\n",
              "      <td>-24.87</td>\n",
              "      <td>...</td>\n",
              "      <td>-13.98</td>\n",
              "      <td>0.62</td>\n",
              "      <td>-5.53</td>\n",
              "      <td>-14.26</td>\n",
              "      <td>-23.69</td>\n",
              "      <td>-37.84</td>\n",
              "      <td>-2.37</td>\n",
              "      <td>6.62</td>\n",
              "      <td>12.02</td>\n",
              "      <td>2.33</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>564</th>\n",
              "      <td>1</td>\n",
              "      <td>144.47</td>\n",
              "      <td>20.09</td>\n",
              "      <td>147.47</td>\n",
              "      <td>109.18</td>\n",
              "      <td>100.87</td>\n",
              "      <td>24.87</td>\n",
              "      <td>115.03</td>\n",
              "      <td>106.37</td>\n",
              "      <td>48.87</td>\n",
              "      <td>...</td>\n",
              "      <td>-57.53</td>\n",
              "      <td>-41.94</td>\n",
              "      <td>-165.94</td>\n",
              "      <td>-265.13</td>\n",
              "      <td>-263.35</td>\n",
              "      <td>-376.53</td>\n",
              "      <td>-176.22</td>\n",
              "      <td>-152.07</td>\n",
              "      <td>-216.60</td>\n",
              "      <td>-202.97</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>565</th>\n",
              "      <td>1</td>\n",
              "      <td>374.46</td>\n",
              "      <td>326.06</td>\n",
              "      <td>319.87</td>\n",
              "      <td>338.23</td>\n",
              "      <td>251.54</td>\n",
              "      <td>209.84</td>\n",
              "      <td>186.35</td>\n",
              "      <td>167.46</td>\n",
              "      <td>135.45</td>\n",
              "      <td>...</td>\n",
              "      <td>-123.55</td>\n",
              "      <td>-166.90</td>\n",
              "      <td>-222.44</td>\n",
              "      <td>-209.71</td>\n",
              "      <td>-180.16</td>\n",
              "      <td>-166.83</td>\n",
              "      <td>-235.66</td>\n",
              "      <td>-213.63</td>\n",
              "      <td>-205.99</td>\n",
              "      <td>-194.07</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>566</th>\n",
              "      <td>1</td>\n",
              "      <td>-0.36</td>\n",
              "      <td>4.96</td>\n",
              "      <td>6.25</td>\n",
              "      <td>4.20</td>\n",
              "      <td>8.26</td>\n",
              "      <td>-9.53</td>\n",
              "      <td>-10.10</td>\n",
              "      <td>-4.54</td>\n",
              "      <td>-11.55</td>\n",
              "      <td>...</td>\n",
              "      <td>-12.40</td>\n",
              "      <td>-5.99</td>\n",
              "      <td>-17.94</td>\n",
              "      <td>-11.96</td>\n",
              "      <td>-12.11</td>\n",
              "      <td>-13.68</td>\n",
              "      <td>-3.59</td>\n",
              "      <td>-5.32</td>\n",
              "      <td>-10.98</td>\n",
              "      <td>-11.24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>567</th>\n",
              "      <td>1</td>\n",
              "      <td>-54.01</td>\n",
              "      <td>-44.13</td>\n",
              "      <td>-41.23</td>\n",
              "      <td>-42.82</td>\n",
              "      <td>-39.47</td>\n",
              "      <td>-24.88</td>\n",
              "      <td>-31.14</td>\n",
              "      <td>-24.71</td>\n",
              "      <td>-13.12</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.73</td>\n",
              "      <td>-1.64</td>\n",
              "      <td>1.58</td>\n",
              "      <td>-4.82</td>\n",
              "      <td>-11.93</td>\n",
              "      <td>-17.14</td>\n",
              "      <td>-4.25</td>\n",
              "      <td>5.47</td>\n",
              "      <td>14.46</td>\n",
              "      <td>18.70</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>568</th>\n",
              "      <td>1</td>\n",
              "      <td>91.36</td>\n",
              "      <td>85.60</td>\n",
              "      <td>48.81</td>\n",
              "      <td>48.69</td>\n",
              "      <td>70.05</td>\n",
              "      <td>22.30</td>\n",
              "      <td>11.63</td>\n",
              "      <td>37.86</td>\n",
              "      <td>28.27</td>\n",
              "      <td>...</td>\n",
              "      <td>2.44</td>\n",
              "      <td>11.53</td>\n",
              "      <td>-16.42</td>\n",
              "      <td>-17.86</td>\n",
              "      <td>21.10</td>\n",
              "      <td>-10.25</td>\n",
              "      <td>-37.06</td>\n",
              "      <td>-8.43</td>\n",
              "      <td>-6.48</td>\n",
              "      <td>17.60</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>569</th>\n",
              "      <td>1</td>\n",
              "      <td>3071.19</td>\n",
              "      <td>2782.53</td>\n",
              "      <td>2608.69</td>\n",
              "      <td>2325.47</td>\n",
              "      <td>2089.37</td>\n",
              "      <td>1769.56</td>\n",
              "      <td>1421.09</td>\n",
              "      <td>1142.09</td>\n",
              "      <td>902.31</td>\n",
              "      <td>...</td>\n",
              "      <td>695.41</td>\n",
              "      <td>865.97</td>\n",
              "      <td>882.41</td>\n",
              "      <td>1203.06</td>\n",
              "      <td>1293.03</td>\n",
              "      <td>1354.41</td>\n",
              "      <td>-192.81</td>\n",
              "      <td>-277.22</td>\n",
              "      <td>-69.63</td>\n",
              "      <td>121.56</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>10 rows × 3198 columns</p>\n",
              "</div>\n",
              "    <div class=\"colab-df-buttons\">\n",
              "\n",
              "  <div class=\"colab-df-container\">\n",
              "    <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-19ed28f1-fca9-4a0f-9dc7-28668b4e7275')\"\n",
              "            title=\"Convert this dataframe to an interactive table.\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\" viewBox=\"0 -960 960 960\">\n",
              "    <path d=\"M120-120v-720h720v720H120Zm60-500h600v-160H180v160Zm220 220h160v-160H400v160Zm0 220h160v-160H400v160ZM180-400h160v-160H180v160Zm440 0h160v-160H620v160ZM180-180h160v-160H180v160Zm440 0h160v-160H620v160Z\"/>\n",
              "  </svg>\n",
              "    </button>\n",
              "\n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    .colab-df-buttons div {\n",
              "      margin-bottom: 4px;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "    <script>\n",
              "      const buttonEl =\n",
              "        document.querySelector('#df-19ed28f1-fca9-4a0f-9dc7-28668b4e7275 button.colab-df-convert');\n",
              "      buttonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "      async function convertToInteractive(key) {\n",
              "        const element = document.querySelector('#df-19ed28f1-fca9-4a0f-9dc7-28668b4e7275');\n",
              "        const dataTable =\n",
              "          await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                    [key], {});\n",
              "        if (!dataTable) return;\n",
              "\n",
              "        const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "          '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "          + ' to learn more about interactive tables.';\n",
              "        element.innerHTML = '';\n",
              "        dataTable['output_type'] = 'display_data';\n",
              "        await google.colab.output.renderOutput(dataTable, element);\n",
              "        const docLink = document.createElement('div');\n",
              "        docLink.innerHTML = docLinkHtml;\n",
              "        element.appendChild(docLink);\n",
              "      }\n",
              "    </script>\n",
              "  </div>\n",
              "\n",
              "\n",
              "<div id=\"df-a6aaea6c-4b07-4b32-b63e-8de559c27fbe\">\n",
              "  <button class=\"colab-df-quickchart\" onclick=\"quickchart('df-a6aaea6c-4b07-4b32-b63e-8de559c27fbe')\"\n",
              "            title=\"Suggest charts\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "<svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "     width=\"24px\">\n",
              "    <g>\n",
              "        <path d=\"M19 3H5c-1.1 0-2 .9-2 2v14c0 1.1.9 2 2 2h14c1.1 0 2-.9 2-2V5c0-1.1-.9-2-2-2zM9 17H7v-7h2v7zm4 0h-2V7h2v10zm4 0h-2v-4h2v4z\"/>\n",
              "    </g>\n",
              "</svg>\n",
              "  </button>\n",
              "\n",
              "<style>\n",
              "  .colab-df-quickchart {\n",
              "      --bg-color: #E8F0FE;\n",
              "      --fill-color: #1967D2;\n",
              "      --hover-bg-color: #E2EBFA;\n",
              "      --hover-fill-color: #174EA6;\n",
              "      --disabled-fill-color: #AAA;\n",
              "      --disabled-bg-color: #DDD;\n",
              "  }\n",
              "\n",
              "  [theme=dark] .colab-df-quickchart {\n",
              "      --bg-color: #3B4455;\n",
              "      --fill-color: #D2E3FC;\n",
              "      --hover-bg-color: #434B5C;\n",
              "      --hover-fill-color: #FFFFFF;\n",
              "      --disabled-bg-color: #3B4455;\n",
              "      --disabled-fill-color: #666;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart {\n",
              "    background-color: var(--bg-color);\n",
              "    border: none;\n",
              "    border-radius: 50%;\n",
              "    cursor: pointer;\n",
              "    display: none;\n",
              "    fill: var(--fill-color);\n",
              "    height: 32px;\n",
              "    padding: 0;\n",
              "    width: 32px;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart:hover {\n",
              "    background-color: var(--hover-bg-color);\n",
              "    box-shadow: 0 1px 2px rgba(60, 64, 67, 0.3), 0 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "    fill: var(--button-hover-fill-color);\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart-complete:disabled,\n",
              "  .colab-df-quickchart-complete:disabled:hover {\n",
              "    background-color: var(--disabled-bg-color);\n",
              "    fill: var(--disabled-fill-color);\n",
              "    box-shadow: none;\n",
              "  }\n",
              "\n",
              "  .colab-df-spinner {\n",
              "    border: 2px solid var(--fill-color);\n",
              "    border-color: transparent;\n",
              "    border-bottom-color: var(--fill-color);\n",
              "    animation:\n",
              "      spin 1s steps(1) infinite;\n",
              "  }\n",
              "\n",
              "  @keyframes spin {\n",
              "    0% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "      border-left-color: var(--fill-color);\n",
              "    }\n",
              "    20% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    30% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    40% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    60% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    80% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "    90% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "  }\n",
              "</style>\n",
              "\n",
              "  <script>\n",
              "    async function quickchart(key) {\n",
              "      const quickchartButtonEl =\n",
              "        document.querySelector('#' + key + ' button');\n",
              "      quickchartButtonEl.disabled = true;  // To prevent multiple clicks.\n",
              "      quickchartButtonEl.classList.add('colab-df-spinner');\n",
              "      try {\n",
              "        const charts = await google.colab.kernel.invokeFunction(\n",
              "            'suggestCharts', [key], {});\n",
              "      } catch (error) {\n",
              "        console.error('Error during call to suggestCharts:', error);\n",
              "      }\n",
              "      quickchartButtonEl.classList.remove('colab-df-spinner');\n",
              "      quickchartButtonEl.classList.add('colab-df-quickchart-complete');\n",
              "    }\n",
              "    (() => {\n",
              "      let quickchartButtonEl =\n",
              "        document.querySelector('#df-a6aaea6c-4b07-4b32-b63e-8de559c27fbe button');\n",
              "      quickchartButtonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "    })();\n",
              "  </script>\n",
              "</div>\n",
              "\n",
              "    </div>\n",
              "  </div>\n"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "dataframe"
            }
          },
          "metadata": {},
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "exoTrain.tail(10)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 412
        },
        "id": "536YnynPGOMI",
        "outputId": "a822bb8a-96e0-4f95-bbfb-6234d95e78b3"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "      LABEL  FLUX.1  FLUX.2  FLUX.3  FLUX.4  FLUX.5  FLUX.6  FLUX.7  FLUX.8  \\\n",
              "5077      1  125.57   78.69   98.29   91.16   78.42   45.82   61.69   22.73   \n",
              "5078      1    7.45   10.02    6.87   -2.82   -1.56   -4.30   -7.01   -6.97   \n",
              "5079      1  475.61  395.50  423.61  376.36  338.94  321.26  326.34  342.84   \n",
              "5080      1  -46.63  -55.39  -64.88  -88.75  -75.40  -64.06  -66.37  -41.95   \n",
              "5081      1  299.41  302.77  278.68  263.48  236.89  186.93  145.45  151.20   \n",
              "5082      1  -91.91  -92.97  -78.76  -97.33  -68.00  -68.24  -75.48  -49.25   \n",
              "5083      1  989.75  891.01  908.53  851.83  755.11  615.78  595.77  458.87   \n",
              "5084      1  273.39  278.00  261.73  236.99  280.73  264.90  252.92  254.88   \n",
              "5085      1    3.82    2.09   -3.29   -2.88    1.66   -0.75    3.85   -0.03   \n",
              "5086      1  323.28  306.36  293.16  287.67  249.89  218.30  188.86  178.93   \n",
              "\n",
              "      FLUX.9  ...  FLUX.3188  FLUX.3189  FLUX.3190  FLUX.3191  FLUX.3192  \\\n",
              "5077   39.09  ...      32.35      63.23      57.98      90.43     115.12   \n",
              "5078   -2.54  ...      -5.25      -8.56       0.53      -4.29      -6.60   \n",
              "5079  251.23  ...     543.25     453.87     344.35     266.16     242.18   \n",
              "5080  -68.07  ...      29.64       6.90      32.94      56.63      28.71   \n",
              "5081  123.38  ...    -126.36    -133.82    -134.02     -98.76    -106.60   \n",
              "5082  -30.92  ...     139.95     147.26     156.95     155.64     156.36   \n",
              "5083  492.84  ...     -26.50      -4.84     -76.30     -37.84    -153.83   \n",
              "5084  237.60  ...     -26.82     -53.89     -48.71      30.99      15.96   \n",
              "5085    3.28  ...      10.86      -3.23      -5.10      -4.61      -9.82   \n",
              "5086  118.93  ...      71.19       0.97      55.20      -1.63      -5.50   \n",
              "\n",
              "      FLUX.3193  FLUX.3194  FLUX.3195  FLUX.3196  FLUX.3197  \n",
              "5077     210.09       3.80      16.33      27.35      21.30  \n",
              "5078       8.75     -10.69      -9.54      -2.48      -8.69  \n",
              "5079     163.02      86.29      13.06     161.22     213.60  \n",
              "5080      28.82     -20.12     -14.41     -43.35     -30.04  \n",
              "5081     -74.95     -46.29      -3.08     -28.43     -48.68  \n",
              "5082     151.75     -24.45     -17.00       3.23      19.28  \n",
              "5083    -136.16      38.03     100.28     -45.64      35.58  \n",
              "5084      -3.47      65.73      88.42      79.07      79.43  \n",
              "5085      -1.50      -4.65     -14.55      -6.41      -2.55  \n",
              "5086     -25.33     -41.31     -16.72     -14.09      27.82  \n",
              "\n",
              "[10 rows x 3198 columns]"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-bc29d05d-2e58-4109-8be7-f9a1bf701f79\" class=\"colab-df-container\">\n",
              "    <div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>LABEL</th>\n",
              "      <th>FLUX.1</th>\n",
              "      <th>FLUX.2</th>\n",
              "      <th>FLUX.3</th>\n",
              "      <th>FLUX.4</th>\n",
              "      <th>FLUX.5</th>\n",
              "      <th>FLUX.6</th>\n",
              "      <th>FLUX.7</th>\n",
              "      <th>FLUX.8</th>\n",
              "      <th>FLUX.9</th>\n",
              "      <th>...</th>\n",
              "      <th>FLUX.3188</th>\n",
              "      <th>FLUX.3189</th>\n",
              "      <th>FLUX.3190</th>\n",
              "      <th>FLUX.3191</th>\n",
              "      <th>FLUX.3192</th>\n",
              "      <th>FLUX.3193</th>\n",
              "      <th>FLUX.3194</th>\n",
              "      <th>FLUX.3195</th>\n",
              "      <th>FLUX.3196</th>\n",
              "      <th>FLUX.3197</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>5077</th>\n",
              "      <td>1</td>\n",
              "      <td>125.57</td>\n",
              "      <td>78.69</td>\n",
              "      <td>98.29</td>\n",
              "      <td>91.16</td>\n",
              "      <td>78.42</td>\n",
              "      <td>45.82</td>\n",
              "      <td>61.69</td>\n",
              "      <td>22.73</td>\n",
              "      <td>39.09</td>\n",
              "      <td>...</td>\n",
              "      <td>32.35</td>\n",
              "      <td>63.23</td>\n",
              "      <td>57.98</td>\n",
              "      <td>90.43</td>\n",
              "      <td>115.12</td>\n",
              "      <td>210.09</td>\n",
              "      <td>3.80</td>\n",
              "      <td>16.33</td>\n",
              "      <td>27.35</td>\n",
              "      <td>21.30</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5078</th>\n",
              "      <td>1</td>\n",
              "      <td>7.45</td>\n",
              "      <td>10.02</td>\n",
              "      <td>6.87</td>\n",
              "      <td>-2.82</td>\n",
              "      <td>-1.56</td>\n",
              "      <td>-4.30</td>\n",
              "      <td>-7.01</td>\n",
              "      <td>-6.97</td>\n",
              "      <td>-2.54</td>\n",
              "      <td>...</td>\n",
              "      <td>-5.25</td>\n",
              "      <td>-8.56</td>\n",
              "      <td>0.53</td>\n",
              "      <td>-4.29</td>\n",
              "      <td>-6.60</td>\n",
              "      <td>8.75</td>\n",
              "      <td>-10.69</td>\n",
              "      <td>-9.54</td>\n",
              "      <td>-2.48</td>\n",
              "      <td>-8.69</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5079</th>\n",
              "      <td>1</td>\n",
              "      <td>475.61</td>\n",
              "      <td>395.50</td>\n",
              "      <td>423.61</td>\n",
              "      <td>376.36</td>\n",
              "      <td>338.94</td>\n",
              "      <td>321.26</td>\n",
              "      <td>326.34</td>\n",
              "      <td>342.84</td>\n",
              "      <td>251.23</td>\n",
              "      <td>...</td>\n",
              "      <td>543.25</td>\n",
              "      <td>453.87</td>\n",
              "      <td>344.35</td>\n",
              "      <td>266.16</td>\n",
              "      <td>242.18</td>\n",
              "      <td>163.02</td>\n",
              "      <td>86.29</td>\n",
              "      <td>13.06</td>\n",
              "      <td>161.22</td>\n",
              "      <td>213.60</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5080</th>\n",
              "      <td>1</td>\n",
              "      <td>-46.63</td>\n",
              "      <td>-55.39</td>\n",
              "      <td>-64.88</td>\n",
              "      <td>-88.75</td>\n",
              "      <td>-75.40</td>\n",
              "      <td>-64.06</td>\n",
              "      <td>-66.37</td>\n",
              "      <td>-41.95</td>\n",
              "      <td>-68.07</td>\n",
              "      <td>...</td>\n",
              "      <td>29.64</td>\n",
              "      <td>6.90</td>\n",
              "      <td>32.94</td>\n",
              "      <td>56.63</td>\n",
              "      <td>28.71</td>\n",
              "      <td>28.82</td>\n",
              "      <td>-20.12</td>\n",
              "      <td>-14.41</td>\n",
              "      <td>-43.35</td>\n",
              "      <td>-30.04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5081</th>\n",
              "      <td>1</td>\n",
              "      <td>299.41</td>\n",
              "      <td>302.77</td>\n",
              "      <td>278.68</td>\n",
              "      <td>263.48</td>\n",
              "      <td>236.89</td>\n",
              "      <td>186.93</td>\n",
              "      <td>145.45</td>\n",
              "      <td>151.20</td>\n",
              "      <td>123.38</td>\n",
              "      <td>...</td>\n",
              "      <td>-126.36</td>\n",
              "      <td>-133.82</td>\n",
              "      <td>-134.02</td>\n",
              "      <td>-98.76</td>\n",
              "      <td>-106.60</td>\n",
              "      <td>-74.95</td>\n",
              "      <td>-46.29</td>\n",
              "      <td>-3.08</td>\n",
              "      <td>-28.43</td>\n",
              "      <td>-48.68</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5082</th>\n",
              "      <td>1</td>\n",
              "      <td>-91.91</td>\n",
              "      <td>-92.97</td>\n",
              "      <td>-78.76</td>\n",
              "      <td>-97.33</td>\n",
              "      <td>-68.00</td>\n",
              "      <td>-68.24</td>\n",
              "      <td>-75.48</td>\n",
              "      <td>-49.25</td>\n",
              "      <td>-30.92</td>\n",
              "      <td>...</td>\n",
              "      <td>139.95</td>\n",
              "      <td>147.26</td>\n",
              "      <td>156.95</td>\n",
              "      <td>155.64</td>\n",
              "      <td>156.36</td>\n",
              "      <td>151.75</td>\n",
              "      <td>-24.45</td>\n",
              "      <td>-17.00</td>\n",
              "      <td>3.23</td>\n",
              "      <td>19.28</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5083</th>\n",
              "      <td>1</td>\n",
              "      <td>989.75</td>\n",
              "      <td>891.01</td>\n",
              "      <td>908.53</td>\n",
              "      <td>851.83</td>\n",
              "      <td>755.11</td>\n",
              "      <td>615.78</td>\n",
              "      <td>595.77</td>\n",
              "      <td>458.87</td>\n",
              "      <td>492.84</td>\n",
              "      <td>...</td>\n",
              "      <td>-26.50</td>\n",
              "      <td>-4.84</td>\n",
              "      <td>-76.30</td>\n",
              "      <td>-37.84</td>\n",
              "      <td>-153.83</td>\n",
              "      <td>-136.16</td>\n",
              "      <td>38.03</td>\n",
              "      <td>100.28</td>\n",
              "      <td>-45.64</td>\n",
              "      <td>35.58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5084</th>\n",
              "      <td>1</td>\n",
              "      <td>273.39</td>\n",
              "      <td>278.00</td>\n",
              "      <td>261.73</td>\n",
              "      <td>236.99</td>\n",
              "      <td>280.73</td>\n",
              "      <td>264.90</td>\n",
              "      <td>252.92</td>\n",
              "      <td>254.88</td>\n",
              "      <td>237.60</td>\n",
              "      <td>...</td>\n",
              "      <td>-26.82</td>\n",
              "      <td>-53.89</td>\n",
              "      <td>-48.71</td>\n",
              "      <td>30.99</td>\n",
              "      <td>15.96</td>\n",
              "      <td>-3.47</td>\n",
              "      <td>65.73</td>\n",
              "      <td>88.42</td>\n",
              "      <td>79.07</td>\n",
              "      <td>79.43</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5085</th>\n",
              "      <td>1</td>\n",
              "      <td>3.82</td>\n",
              "      <td>2.09</td>\n",
              "      <td>-3.29</td>\n",
              "      <td>-2.88</td>\n",
              "      <td>1.66</td>\n",
              "      <td>-0.75</td>\n",
              "      <td>3.85</td>\n",
              "      <td>-0.03</td>\n",
              "      <td>3.28</td>\n",
              "      <td>...</td>\n",
              "      <td>10.86</td>\n",
              "      <td>-3.23</td>\n",
              "      <td>-5.10</td>\n",
              "      <td>-4.61</td>\n",
              "      <td>-9.82</td>\n",
              "      <td>-1.50</td>\n",
              "      <td>-4.65</td>\n",
              "      <td>-14.55</td>\n",
              "      <td>-6.41</td>\n",
              "      <td>-2.55</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5086</th>\n",
              "      <td>1</td>\n",
              "      <td>323.28</td>\n",
              "      <td>306.36</td>\n",
              "      <td>293.16</td>\n",
              "      <td>287.67</td>\n",
              "      <td>249.89</td>\n",
              "      <td>218.30</td>\n",
              "      <td>188.86</td>\n",
              "      <td>178.93</td>\n",
              "      <td>118.93</td>\n",
              "      <td>...</td>\n",
              "      <td>71.19</td>\n",
              "      <td>0.97</td>\n",
              "      <td>55.20</td>\n",
              "      <td>-1.63</td>\n",
              "      <td>-5.50</td>\n",
              "      <td>-25.33</td>\n",
              "      <td>-41.31</td>\n",
              "      <td>-16.72</td>\n",
              "      <td>-14.09</td>\n",
              "      <td>27.82</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>10 rows × 3198 columns</p>\n",
              "</div>\n",
              "    <div class=\"colab-df-buttons\">\n",
              "\n",
              "  <div class=\"colab-df-container\">\n",
              "    <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-bc29d05d-2e58-4109-8be7-f9a1bf701f79')\"\n",
              "            title=\"Convert this dataframe to an interactive table.\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\" viewBox=\"0 -960 960 960\">\n",
              "    <path d=\"M120-120v-720h720v720H120Zm60-500h600v-160H180v160Zm220 220h160v-160H400v160Zm0 220h160v-160H400v160ZM180-400h160v-160H180v160Zm440 0h160v-160H620v160ZM180-180h160v-160H180v160Zm440 0h160v-160H620v160Z\"/>\n",
              "  </svg>\n",
              "    </button>\n",
              "\n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    .colab-df-buttons div {\n",
              "      margin-bottom: 4px;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "    <script>\n",
              "      const buttonEl =\n",
              "        document.querySelector('#df-bc29d05d-2e58-4109-8be7-f9a1bf701f79 button.colab-df-convert');\n",
              "      buttonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "      async function convertToInteractive(key) {\n",
              "        const element = document.querySelector('#df-bc29d05d-2e58-4109-8be7-f9a1bf701f79');\n",
              "        const dataTable =\n",
              "          await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                    [key], {});\n",
              "        if (!dataTable) return;\n",
              "\n",
              "        const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "          '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "          + ' to learn more about interactive tables.';\n",
              "        element.innerHTML = '';\n",
              "        dataTable['output_type'] = 'display_data';\n",
              "        await google.colab.output.renderOutput(dataTable, element);\n",
              "        const docLink = document.createElement('div');\n",
              "        docLink.innerHTML = docLinkHtml;\n",
              "        element.appendChild(docLink);\n",
              "      }\n",
              "    </script>\n",
              "  </div>\n",
              "\n",
              "\n",
              "<div id=\"df-ac9d30b9-0c54-4b64-9225-0da919ae0bd1\">\n",
              "  <button class=\"colab-df-quickchart\" onclick=\"quickchart('df-ac9d30b9-0c54-4b64-9225-0da919ae0bd1')\"\n",
              "            title=\"Suggest charts\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "<svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "     width=\"24px\">\n",
              "    <g>\n",
              "        <path d=\"M19 3H5c-1.1 0-2 .9-2 2v14c0 1.1.9 2 2 2h14c1.1 0 2-.9 2-2V5c0-1.1-.9-2-2-2zM9 17H7v-7h2v7zm4 0h-2V7h2v10zm4 0h-2v-4h2v4z\"/>\n",
              "    </g>\n",
              "</svg>\n",
              "  </button>\n",
              "\n",
              "<style>\n",
              "  .colab-df-quickchart {\n",
              "      --bg-color: #E8F0FE;\n",
              "      --fill-color: #1967D2;\n",
              "      --hover-bg-color: #E2EBFA;\n",
              "      --hover-fill-color: #174EA6;\n",
              "      --disabled-fill-color: #AAA;\n",
              "      --disabled-bg-color: #DDD;\n",
              "  }\n",
              "\n",
              "  [theme=dark] .colab-df-quickchart {\n",
              "      --bg-color: #3B4455;\n",
              "      --fill-color: #D2E3FC;\n",
              "      --hover-bg-color: #434B5C;\n",
              "      --hover-fill-color: #FFFFFF;\n",
              "      --disabled-bg-color: #3B4455;\n",
              "      --disabled-fill-color: #666;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart {\n",
              "    background-color: var(--bg-color);\n",
              "    border: none;\n",
              "    border-radius: 50%;\n",
              "    cursor: pointer;\n",
              "    display: none;\n",
              "    fill: var(--fill-color);\n",
              "    height: 32px;\n",
              "    padding: 0;\n",
              "    width: 32px;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart:hover {\n",
              "    background-color: var(--hover-bg-color);\n",
              "    box-shadow: 0 1px 2px rgba(60, 64, 67, 0.3), 0 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "    fill: var(--button-hover-fill-color);\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart-complete:disabled,\n",
              "  .colab-df-quickchart-complete:disabled:hover {\n",
              "    background-color: var(--disabled-bg-color);\n",
              "    fill: var(--disabled-fill-color);\n",
              "    box-shadow: none;\n",
              "  }\n",
              "\n",
              "  .colab-df-spinner {\n",
              "    border: 2px solid var(--fill-color);\n",
              "    border-color: transparent;\n",
              "    border-bottom-color: var(--fill-color);\n",
              "    animation:\n",
              "      spin 1s steps(1) infinite;\n",
              "  }\n",
              "\n",
              "  @keyframes spin {\n",
              "    0% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "      border-left-color: var(--fill-color);\n",
              "    }\n",
              "    20% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    30% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    40% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    60% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    80% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "    90% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "  }\n",
              "</style>\n",
              "\n",
              "  <script>\n",
              "    async function quickchart(key) {\n",
              "      const quickchartButtonEl =\n",
              "        document.querySelector('#' + key + ' button');\n",
              "      quickchartButtonEl.disabled = true;  // To prevent multiple clicks.\n",
              "      quickchartButtonEl.classList.add('colab-df-spinner');\n",
              "      try {\n",
              "        const charts = await google.colab.kernel.invokeFunction(\n",
              "            'suggestCharts', [key], {});\n",
              "      } catch (error) {\n",
              "        console.error('Error during call to suggestCharts:', error);\n",
              "      }\n",
              "      quickchartButtonEl.classList.remove('colab-df-spinner');\n",
              "      quickchartButtonEl.classList.add('colab-df-quickchart-complete');\n",
              "    }\n",
              "    (() => {\n",
              "      let quickchartButtonEl =\n",
              "        document.querySelector('#df-ac9d30b9-0c54-4b64-9225-0da919ae0bd1 button');\n",
              "      quickchartButtonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "    })();\n",
              "  </script>\n",
              "</div>\n",
              "\n",
              "    </div>\n",
              "  </div>\n"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "dataframe"
            }
          },
          "metadata": {},
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "print(exoTest['LABEL'].value_counts())\n",
        "print(exoTrain['LABEL'].value_counts())"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "aCoGDTHDL3ZM",
        "outputId": "d6ffbf9f-10c8-4e36-9a01-2b0936e877f4"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "1    565\n",
            "2      5\n",
            "Name: LABEL, dtype: int64\n",
            "1    5050\n",
            "2      37\n",
            "Name: LABEL, dtype: int64\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "> **Key takeways:**\n",
        "\n",
        "- You should notice the large imbalance in those datasets. Because of that, we will apply SMOTE to regularize the data later on in this notebook.\n",
        "\n",
        "- From the print function we could see that, in both datasets, the number of positive samples is overwhelmingly smaller than the negative ones. Soon we will further analyze the disparity between samples in the training dataset, which will be our main focus.\n",
        "\n",
        "\n"
      ],
      "metadata": {
        "id": "VGV1T8CQJyEr"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Hooray! Now that the setup is done, let's customize the dataset to our needs. We'll also augment the data to adress the imbalance issue.\n",
        "\n"
      ],
      "metadata": {
        "id": "uu3zytZFC0Wo"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "---"
      ],
      "metadata": {
        "id": "bCYqmySdQDco"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Data customization and augmentation"
      ],
      "metadata": {
        "id": "r9XA0r-eDYIM"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "#### **Customization**"
      ],
      "metadata": {
        "id": "SlYWjrcueNyZ"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "- Continuing on the data preprocessing, we will use PyTorch to custom the data."
      ],
      "metadata": {
        "id": "ijKQPHpOQzsp"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "class CustomDataset(Dataset):\n",
        "    def __init__(self, csv_file, transform=None):\n",
        "        with open(csv_file,\"r\") as f:\n",
        "            self.data = pd.read_csv(f) # read the files\n",
        "        self.labels = self.data.iloc[:,0].values - 1 # set the first line of the input data as the label (Originally 1 or 2, but we will transform them into 0 or 1 by decreasing 1)\n",
        "        self.features = self.data.iloc[:, 1:].values # set the rest of the input data as the feature (FLUX over time)\n",
        "        self.transform = transform # transformation (which is None) that will be applied to samples.\n",
        "\n",
        "    def __len__(self):\n",
        "        return len(self.data)\n",
        "\n",
        "    def __getitem__(self, idx): # retrieves a data sample from the dataset\n",
        "            label = self.labels[idx] # fetch label of sample\n",
        "            feature = self.features[idx] # fetch features of sample\n",
        "\n",
        "            if self.transform: # if there is a specified transformation, transform the data\n",
        "                feature = self.transform(feature)\n",
        "\n",
        "            sample = {'feature': feature, 'label': label}\n",
        "            return sample\n",
        "\n",
        "# instantiate train and test datasets\n",
        "train_dataset = CustomDataset('./exoTrain.csv') # grab the training data\n",
        "test_dataset = CustomDataset('./exoTest.csv') # grab the test data\n"
      ],
      "metadata": {
        "id": "sXE-TFntRveA"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        " Following [PyTorch's official tutorial for deep learning frameworks](https://pytorch.org/tutorials/beginner/data_loading_tutorial.html), we have the CSV file opened and its contents read using `pd.read_csv()` from the Pandas library.\n",
        "\n",
        " The data is stored in the `self.data` attribute. The labels are extracted from the first column `(iloc[:,0])` of the data using `.values`.\n",
        "\n",
        " The labels are then transformed by subtracting 1, converting them from 1 or 2 to 0 or 1 respectively. These transformed labels are stored in the `self.labels` attribute."
      ],
      "metadata": {
        "id": "CGKdI5IiUQJQ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "    def __len__(self): # function that gives back the size of the dataset (how many samples)\n",
        "            return len(self.labels)\n",
        "\n",
        "    def __getitem__(self, idx): # retrieves a data sample from the dataset\n",
        "            label = self.labels[idx] # fetch label of sample\n",
        "            feature = self.features[idx] # fetch features of sample\n",
        "\n",
        "            if self.transform: # if there is a specified transformation, transform the data\n",
        "                feature = self.transform(feature)\n",
        "\n",
        "            sample = {'feature': feature, 'label': label}\n",
        "            return sample\n",
        "\n",
        "    train_dataset = CustomDataset('./exoTrain.csv') # grab the training data\n",
        "    test_dataset = CustomDataset('./exoTest.csv') # grab the test data\n",
        "\n",
        "\n"
      ],
      "metadata": {
        "id": "Rm-P3f0WXO2u"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "\n",
        "\n",
        "To check if the transformation was successful, you can call a print function as we did before, but this time adding `pd.Series`."
      ],
      "metadata": {
        "id": "aZM3RgDKW0iw"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(\"LABEL:\", pd.Series(test_dataset.labels).value_counts())\n",
        "print(\"LABEL:\", pd.Series(train_dataset.labels).value_counts())"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "w-vIEvdNZUno",
        "outputId": "1c369e19-8253-4ac1-837b-96060444641f"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "LABEL: 0    565\n",
            "1      5\n",
            "dtype: int64\n",
            "LABEL: 0    5050\n",
            "1      37\n",
            "dtype: int64\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "label_counts = np.bincount(train_dataset.labels)\n",
        "label_names = ['Not Exoplanet','Exoplanet']\n",
        "\n",
        "fig, axes = plt.subplots(1, 2, figsize=(12, 6))\n",
        "\n",
        "for i, dataset in enumerate([train_dataset, test_dataset]):\n",
        "    label_counts = np.bincount(dataset.labels)\n",
        "    axes[i].pie(label_counts, labels=label_names, autopct='%1.1f%%', startangle=90, colors=sns.color_palette('pastel'))\n",
        "    axes[i].set_title('Distribution of Samples in ' + ['Training', 'Test'][i] + ' Dataset')\n",
        "\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 417
        },
        "id": "kg-FFLpBcF2N",
        "outputId": "14e5ddc2-08ff-4aee-ba30-c3e5ed4052f7"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1200x600 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "It's evident that the training dataset exhibits slightly more imbalance compared to the test dataset."
      ],
      "metadata": {
        "id": "bWDSyr1TcZOj"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "#### **Augmentation**"
      ],
      "metadata": {
        "id": "IJwMknnTeDrY"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "According to [Chawla, Bowyer, et al. (2002)](https://www.jair.org/index.php/jair/article/view/10302), Synthetic Minority Over-Sampling Technique (SMOTE) combines under-sampling of the majority class -- in our case, the negative samples of exoplanet candidates -- with a unique form of over-sampling the minority class to address class imbalance."
      ],
      "metadata": {
        "id": "GBQAQhgzjJU3"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "So, in essence, SMOTE balances the samples by creating synthetic data points from the minority class."
      ],
      "metadata": {
        "id": "ndMPcB2UeI1K"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "\n",
        "\n",
        "# initialize smote with default sampling strategy\n",
        "smote = SMOTE()\n",
        "\n",
        "\n",
        "# update the labels and features to the resampled data\n",
        "train_dataset.features, train_dataset.labels = smote.fit_resample(train_dataset.features, train_dataset.labels)\n",
        "\n",
        "\n",
        "print(\"Class distribution in the testing dataset after SMOTE:\", pd.Series(test_dataset.labels).value_counts())\n",
        "print(\"Class distribution in the training dataset after SMOTE:\", pd.Series(train_dataset.labels).value_counts())"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "p4Fr6CMjfs4z",
        "outputId": "8a8a6e93-429a-474c-e253-65c22fbfa0dd"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Class distribution in the testing dataset after SMOTE: 0    565\n",
            "1      5\n",
            "dtype: int64\n",
            "Class distribution in the training dataset after SMOTE: 1    5050\n",
            "0    5050\n",
            "dtype: int64\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "label_counts = np.bincount(train_dataset.labels)\n",
        "label_names = ['Not Exoplanet','Exoplanet']\n",
        "\n",
        "fig, axes = plt.subplots(1, 2, figsize=(12, 6))\n",
        "\n",
        "for i, dataset in enumerate([train_dataset, test_dataset]):\n",
        "    label_counts = np.bincount(dataset.labels)\n",
        "    axes[i].pie(label_counts, labels=label_names, autopct='%1.1f%%', startangle=90, colors=sns.color_palette('pastel'))\n",
        "    axes[i].set_title('Distribution of Samples in ' + ['Training', 'Test'][i] + ' Dataset')\n",
        "\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 389
        },
        "id": "I8I4Ou7sq5o8",
        "outputId": "bd96a669-01aa-418f-b7cd-70f05e8ad2db"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1200x600 with 2 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAA/4AAAHRCAYAAADT+WIHAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAACE/klEQVR4nO3dd3wUdf7H8ffsbnbTGy3UhN4EFFFEEMFy2PXOhhXs51k5u54FsDcs589+oOjZFT3sDQuggiLSO6EmQBqpW+f3R2QlEiAJSWZ39vV8PPYBOzu7+5nN7sz3862GaZqmAAAAAACALTmsDgAAAAAAADQdEn8AAAAAAGyMxB8AAAAAABsj8QcAAAAAwMZI/AEAAAAAsDESfwAAAAAAbIzEHwAAAAAAGyPxBwAAAADAxkj8AQAAAACwMRJ/SJLuuusuGYbRLO81YsQIjRgxInx/xowZMgxDb7/9drO8/9ixY5WTk9Ms79VQZWVluvjii5WVlSXDMHTttddaHVKTaM7v3Q7R8Pf/s335nKZMmSLDMLR27drGDQpATKK8EFkoLzSdaPj7A/VB4m9DOwr6O27x8fFq166dRo0apSeeeEKlpaWN8j6bNm3SXXfdpV9//bVRXq8xRXJsdXHvvfdqypQpuvzyyzV16lSdd955u93X5/Pp8ccf1wEHHKDU1FSlp6erb9++uvTSS7V06dJmjDr25OTk1Pit7e42ZcoUq0O1xI6C2o5bYmKiOnXqpBNPPFGTJ0+W1+tt8Gt/9NFHuuuuuxov2H107733atq0aVaHAdQL5YXIjq0uKC9Eh+YuL9TnmrR27doaMcTFxally5Y69NBDdeutt2rdunUNjiPSfl+RVnZoboZpmqbVQaBxTZkyRRdccIEmTJigzp07y+/3Ky8vTzNmzNDnn3+uTp066YMPPlD//v3DzwkEAgoEAoqPj6/z+8ydO1cHHXSQJk+erLFjx9b5eT6fT5LkdrslVdfgjxw5Um+99ZZOO+20Or9OQ2Pz+/0KhULyeDyN8l5N4ZBDDpHL5dL333+/131PPPFEffzxxzrrrLM0ZMgQ+f1+LV26VNOnT9fEiRPr9bdpbnfddZfGjx+v5jwNNebff9q0aSorKwvf/+ijj/Taa69p0qRJatmyZXj7oYceqi5dujT4fRry+9whGAzK7/fL4/E0e2vJjr/v008/reTkZHm9Xm3cuFGffvqpZs2apf79+2v69Onq2LFjvV/7yiuv1FNPPdWs3509SU5O1mmnnRazlTyITpQXKC9QXti9aCwv7FCfa9LatWvVuXNnnXXWWTruuOMUCoVUVFSkOXPm6N1335VhGHrxxRc1evToesfR0N9+U4m0skNzc1kdAJrOscceq0GDBoXv33LLLfrqq690wgkn6KSTTtKSJUuUkJAgSXK5XHK5mvbrUFFRocTExPAF3CpxcXGWvn9dbNmyRX369NnrfnPmzNH06dN1zz336NZbb63x2L///W8VFxc3UYTRqzH//qecckqN+3l5eXrttdd0yimn7LF7YHl5uZKSkur8Pvvy+3Q6nXI6nQ16bmM57bTTahRs7rjjDr366qs6//zzdfrpp+uHH36wMDoAlBdqR3khtkVCeaE5DRw4UOeee26Nbbm5ufrLX/6iMWPGqHfv3howYIBF0aEx0NU/xhxxxBG6/fbblZubq1deeSW8vbaxU59//rmGDRum9PR0JScnq2fPnuGLxYwZM3TQQQdJki644IJduiiNGDFC++23n37++WcNHz5ciYmJ4ef+eczeDsFgULfeequysrKUlJSkk046SevXr6+xT05OTq01hju/5t5iq23MVnl5ua677jp17NhRHo9HPXv21MMPP7xLjaBhGLryyis1bdo07bfffvJ4POrbt68++eST2j/wP9myZYsuuugitWnTRvHx8RowYIBeeuml8OM7xi+uWbNGH374YTj23Y3PXrVqlSRp6NChuzzmdDrVokWL8P3c3Fz94x//UM+ePZWQkKAWLVro9NNP3+W1d3T9/P7773X11VerVatWSk9P12WXXSafz6fi4mKdf/75ysjIUEZGhm688cYan9OOLmMPP/ywJk2apOzsbCUkJOjwww/XwoUL6/Q5vfLKKzrwwAOVkJCgzMxMjR49epfvwooVK3TqqacqKytL8fHx6tChg0aPHq2SkpI9vvaf//47x/vcc8+pa9eu8ng8OuiggzRnzpw6xbu390tOTtaqVat03HHHKSUlReecc44k6bvvvtPpp5+uTp06yePxqGPHjho3bpwqKytrvEZtv8+6fhdrG+Ofk5OjE044Qd9//70OPvhgxcfHq0uXLnr55Zd3if+3337T4YcfroSEBHXo0EF33323Jk+evM/zBpxzzjm6+OKL9eOPP+rzzz8Pb6/LZzJ27Fg99dRT4c9hx22Hhx9+WIceeqhatGihhIQEHXjggbWOCd7TOW4Hr9erO++8U926dQvHc+ONN9YYpmAYhsrLy/XSSy+FY4mElg1gX1BeoLxAeaF5ywuNdTyNeU3Kzs7WlClT5PP59OCDD4a3FxYW6vrrr1e/fv2UnJys1NRUHXvssZo/f354n739vupaBsrLy9MFF1ygDh06yOPxqG3btjr55JN3+T5+/PHHOuyww5SUlKSUlBQdf/zxWrRoUfjxvZUdYgEt/jHovPPO06233qrPPvtMl1xySa37LFq0SCeccIL69++vCRMmyOPxaOXKlZo5c6YkqXfv3powYYLuuOMOXXrppTrssMMkVXdR2qGgoEDHHnusRo8erXPPPVdt2rTZY1z33HOPDMPQTTfdpC1btuixxx7TUUcdpV9//TXc0lAXdYltZ6Zp6qSTTtLXX3+tiy66SPvvv78+/fRT3XDDDdq4caMmTZpUY//vv/9e7777rv7xj38oJSVFTzzxhE499VStW7euxoXzzyorKzVixAitXLlSV155pTp37qy33npLY8eOVXFxsa655hr17t1bU6dO1bhx49ShQwddd911kqRWrVrV+prZ2dmSpFdffVVDhw7dYyvMnDlzNGvWLI0ePVodOnTQ2rVr9fTTT2vEiBFavHixEhMTa+x/1VVXKSsrS+PHj9cPP/yg5557Tunp6Zo1a5Y6deqke++9Vx999JEeeugh7bfffjr//PNrPP/ll19WaWmprrjiClVVVenxxx/XEUccoQULFuzxu3DPPffo9ttv1xlnnKGLL75YW7du1ZNPPqnhw4dr3rx5Sk9Pl8/n06hRo+T1esNxbty4UdOnT1dxcbHS0tJ2+/q789///lelpaW67LLLZBiGHnzwQf3tb3/T6tWr97nWPxAIaNSoURo2bJgefvjh8Gf91ltvqaKiQpdffrlatGihn376SU8++aQ2bNigt956a6+v29DvoiStXLlSp512mi666CKNGTNG//nPfzR27FgdeOCB6tu3ryRp48aNGjlypAzD0C233KKkpCS98MILjdbt9bzzztNzzz2nzz77TEcffbSkun0ml112mTZt2qTPP/9cU6dO3eV1H3/8cZ100kk655xz5PP59Prrr+v000/X9OnTdfzxx0va+zlOkkKhkE466SR9//33uvTSS9W7d28tWLBAkyZN0vLly8PjJ6dOnaqLL75YBx98sC699FJJUteuXRvlMwKsRHmhJsoLlBekpisvNNbxNPY1aciQIeratWuNSvrVq1dr2rRpOv3009W5c2fl5+fr2Wef1eGHH67FixerXbt2e/191bUMdOqpp2rRokW66qqrlJOToy1btujzzz/XunXrwhUzU6dO1ZgxYzRq1Cg98MADqqio0NNPP61hw4Zp3rx5ysnJ2WvZISaYsJ3Jkyebksw5c+bsdp+0tDTzgAMOCN+/8847zZ2/DpMmTTIlmVu3bt3ta8yZM8eUZE6ePHmXxw4//HBTkvnMM8/U+tjhhx8evv/111+bksz27dub27dvD29/8803TUnm448/Ht6WnZ1tjhkzZq+vuafYxowZY2ZnZ4fvT5s2zZRk3n333TX2O+2000zDMMyVK1eGt0ky3W53jW3z5883JZlPPvnkLu+1s8cee8yUZL7yyivhbT6fzxwyZIiZnJxc49izs7PN448/fo+vZ5qmGQqFwp91mzZtzLPOOst86qmnzNzc3F32raio2GXb7NmzTUnmyy+/HN624/szatQoMxQKhbcPGTLENAzD/Pvf/x7eFggEzA4dOtT47NesWWNKMhMSEswNGzaEt//444+mJHPcuHHhbX/+3q1du9Z0Op3mPffcUyPOBQsWmC6XK7x93rx5piTzrbfe2utn9Gd//vvviLdFixZmYWFhePv7779vSjL/97//1fm1H3roIVOSuWbNmhrvJ8m8+eabd9m/tr/JfffdZxqGUeNv+OfPyTTr/l3c8ffcOabs7GxTkvntt9+Gt23ZssX0eDzmddddF9521VVXmYZhmPPmzQtvKygoMDMzM3d5zdrsiHt355GioiJTkvnXv/41vK2un8kVV1yxy2eyu9fw+XzmfvvtZx5xxBHhbXU5x02dOtV0OBzmd999V2P7M888Y0oyZ86cGd6WlJRU67kJiGSUFygvUF7YveYsLzT28dTnmrTjuB566KHd7nPyySebksySkhLTNE2zqqrKDAaDu7yOx+MxJ0yYEN62p99XXa73O8oJe4qttLTUTE9PNy+55JIa2/Py8sy0tLQa2/dUdogFdPWPUcnJyXucrTc9PV2S9P777ysUCjXoPTwejy644II673/++ecrJSUlfP+0005T27Zt9dFHHzXo/evqo48+ktPp1NVXX11j+3XXXSfTNPXxxx/X2H7UUUfVqDnt37+/UlNTtXr16r2+T1ZWls4666zwtri4OF199dUqKyvTN998U+/YDcPQp59+qrvvvlsZGRl67bXXdMUVVyg7O1tnnnlmjTF7O7eC+P1+FRQUqFu3bkpPT9cvv/yyy2tfdNFFNbpADR48WKZp6qKLLgpvczqdGjRoUK3Hfsopp6h9+/bh+wcffLAGDx68x7/nu+++q1AopDPOOEPbtm0L37KystS9e3d9/fXXkhSuof/0009VUVFRh09q784880xlZGSE7++omd7b37WuLr/88l227fw3KS8v17Zt23TooYfKNE3Nmzdvr6/Z0O+iJPXp0yd8jFJ1K1HPnj1rPPeTTz7RkCFDtP/++4e3ZWZmhocq7Kvk5GRJqnEu2tfP5M+vUVRUpJKSEh122GE1vud1Oce99dZb6t27t3r16lXj+3jEEUdIUvj7CNgZ5YU/UF6gvCA1TXnByuOpiz9frz0ejxyO6jQyGAyqoKAgPMyntu9IbepyvU9ISJDb7daMGTNUVFRU6+t8/vnnKi4u1llnnVXjs3M6nRo8eDDX6p2Q+MeosrKyGhfNPzvzzDM1dOhQXXzxxWrTpo1Gjx6tN998s14X9fbt29drYp7u3bvXuG8Yhrp169bk64/n5uaqXbt2u3wevXv3Dj++s06dOu3yGhkZGbs9Ie38Pt27dw+fKPf2PnXl8Xh02223acmSJdq0aZNee+01HXLIIXrzzTd15ZVXhverrKzUHXfcER6X2LJlS7Vq1UrFxcW1jnP783HuuNj8eQb2tLS0Wo/9z39PSerRo8ce/54rVqyQaZrq3r27WrVqVeO2ZMkSbdmyRZLUuXNn/fOf/9QLL7ygli1batSoUXrqqaf2Ol5vT/58vDsu6nv7u9aFy+VShw4ddtm+bt06jR07VpmZmUpOTlarVq10+OGHS1KdjqWh38W6Pjc3N1fdunXbZb/atjXEjhmOd/7t7etnIknTp0/XIYccovj4eGVmZqpVq1Z6+umnazy/Lue4FStWaNGiRbt8F3v06CFJ4e8jYGeUF/5AeYHygtQ05QUrj6cu/ny9DoVCmjRpkrp3717jO/Lbb7/VOZa6XO89Ho8eeOABffzxx2rTpo2GDx+uBx98UHl5eeHXWbFihaTqeUn+/Nl99tlnXKt3whj/GLRhwwaVlJTssfCekJCgb7/9Vl9//bU+/PBDffLJJ3rjjTd0xBFH6LPPPqvTLOH1GWdXV7ubhCMYDDbbzOW7ex8zApYGadu2rUaPHq1TTz1Vffv21ZtvvqkpU6bI5XLpqquu0uTJk3XttddqyJAhSktLk2EYGj16dK0FtN0dZ23bG+vYQ6GQDMPQxx9/XOv77KhxlqRHHnlEY8eO1fvvv6/PPvtMV199te677z798MMPtSbZe9OUf9eda8Z3CAaDOvroo1VYWKibbrpJvXr1UlJSkjZu3KixY8fWqdC8LzFHwvd4x+RNO85FjfGZfPfddzrppJM0fPhw/d///Z/atm2ruLg4TZ48Wf/973/D+9XlHBcKhdSvXz89+uijtb5XQ5YhBKIJ5YV9Ewnn2d2hvBBZ5QUrj6cuFi5cqNatWys1NVWSdO+99+r222/XhRdeqIkTJyozM1MOh0PXXnttna7V9bneX3vttTrxxBM1bdo0ffrpp7r99tt133336auvvtIBBxwQ3nfq1KnKysra5b2aehWSaMInEYN2TGgxatSoPe7ncDh05JFH6sgjj9Sjjz6qe++9V7fddpu+/vprHXXUUY0+E+aOGrsdTNPUypUra6wfnJGRUeuSM7m5uTXWPa1PbNnZ2friiy9UWlpaoxZ/6dKl4ccbQ3Z2tn777TeFQqEaSWBjv49U3SWwf//+WrFiRbir2Ntvv60xY8bokUceCe9XVVXVZEv4/PnvKUnLly/f47I1Xbt2lWma6ty5c7hVdU/69eunfv366V//+pdmzZqloUOH6plnntHdd9+9L6E3iwULFmj58uV66aWXakx0tPPkOVbLzs7WypUrd9le27aG+PO5qD6fye5+4++8847i4+P16aef1piEcPLkybvsu7dzXNeuXTV//nwdeeSRez2nxNrMwIgNlBdqorxQ3GjvuzPKC41/PI35m5s9e7ZWrVpVY6m/t99+WyNHjtSLL75YY9/i4uIay/fuLo76loG6du2q6667Ttddd51WrFih/fffX4888oheeeWV8HCa1q1b66ijjtrjscT6tZqu/jHmq6++0sSJE9W5c+c9jtMtLCzcZduOcb47lrHasQ55Y10IdszqusPbb7+tzZs369hjjw1v69q1q3744Qf5fL7wtunTp++y1El9YjvuuOMUDAb173//u8b2SZMmyTCMGu+/L4477jjl5eXpjTfeCG8LBAJ68sknlZycHO7eVB8rVqzQunXrdtleXFys2bNnKyMjIzzDr9Pp3KU2+sknn1QwGKz3+9bFtGnTtHHjxvD9n376ST/++OMeP8+//e1vcjqdGj9+/C6xmqapgoICSdL27dsVCARqPN6vXz85HI4ay6xFsh01+jsfp2maevzxx60KaRejRo3S7Nmz9euvv4a3FRYW6tVXX93n1/7vf/+rF154QUOGDNGRRx4pqX6fye5+406nU4Zh1Pher127NjwD/87H8Wd/PsedccYZ2rhxo55//vld9q2srFR5eXmNeFgHG3ZCeWFXlBcoLzSVxj6exrom5ebmauzYsXK73brhhhvC22v7jrz11ls1/o474pBqv1ZLe7/eV1RUqKqqqsa2rl27KiUlJXy8o0aNUmpqqu699175/f5djmHr1q17jSdW0OJvYx9//LGWLl2qQCCg/Px8ffXVV/r888+VnZ2tDz74QPHx8bt97oQJE/Ttt9/q+OOPV3Z2trZs2aL/+7//U4cOHTRs2DBJ1T+89PR0PfPMM0pJSVFSUpIGDx6szp07NyjezMxMDRs2TBdccIHy8/P12GOPqVu3bjWWELr44ov19ttv65hjjtEZZ5yhVatW1ajt26E+sZ144okaOXKkbrvtNq1du1YDBgzQZ599pvfff1/XXnttoy3Ldemll+rZZ5/V2LFj9fPPPysnJ0dvv/22Zs6cqccee2yPYyh3Z/78+Tr77LN17LHH6rDDDlNmZqY2btyol156SZs2bdJjjz0WPrmecMIJmjp1qtLS0tSnTx/Nnj1bX3zxxV6XfWuobt26adiwYbr88svl9Xr12GOPqUWLFrrxxht3+5yuXbvq7rvv1i233KK1a9fqlFNOUUpKitasWaP33ntPl156qa6//np99dVXuvLKK3X66aerR48eCgQCmjp1qpxOp0499dQmOZ7G1qtXL3Xt2lXXX3+9Nm7cqNTUVL3zzjuNMqdAY7nxxhv1yiuv6Oijj9ZVV10VXs6vU6dOKiwsrHPN+dtvv63k5GT5fD5t3LhRn376qWbOnKkBAwbUWLKnPp/JgQceKEm6+uqrNWrUKDmdTo0ePVrHH3+8Hn30UR1zzDE6++yztWXLFj311FPq1q2bfvvtt/Dz63KOO++88/Tmm2/q73//u77++msNHTpUwWBQS5cu1ZtvvqlPP/1UgwYNCsfzxRdf6NFHH1W7du3UuXNnDR48uMGfPdCcKC9QXqC8YK3GPp6GXJN++eUXvfLKKwqFQiouLtacOXP0zjvvyDAMTZ06tUaPmhNOOEETJkzQBRdcoEMPPVQLFizQq6++WqM3zY7jqu33Vdfr/fLly3XkkUfqjDPOUJ8+feRyufTee+8pPz9fo0ePliSlpqbq6aef1nnnnaeBAwdq9OjRatWqldatW6cPP/xQQ4cODVfW7a7sEDOacskAWGPH8io7bm6328zKyjKPPvpo8/HHH6+xDMwOf14m5csvvzRPPvlks127dqbb7TbbtWtnnnXWWeby5ctrPO/99983+/TpY7pcrhrLdRx++OFm3759a41vd8vzvPbaa+Ytt9xitm7d2kxISDCPP/74WpeZeeSRR8z27dubHo/HHDp0qDl37txdXnNPsf15eRbTrF4KZNy4cWa7du3MuLg4s3v37uZDDz1UY3ka06xenueKK67YJabdLRv0Z/n5+eYFF1xgtmzZ0nS73Wa/fv1qXeKkrsvz5Ofnm/fff795+OGHm23btjVdLpeZkZFhHnHEEebbb79dY9+ioqLweycnJ5ujRo0yly5dukvsu1veaXdLs40ZM8ZMSkoK3995WZhHHnnE7Nixo+nxeMzDDjvMnD9/fq2v+WfvvPOOOWzYMDMpKclMSkoye/XqZV5xxRXmsmXLTNM0zdWrV5sXXnih2bVrVzM+Pt7MzMw0R44caX7xxRd7/cx2tzxPbUvFSDLvvPPOvb7mDrtbzm/nz2dnixcvNo866igzOTnZbNmypXnJJZeEl3va+Xuxu+X86vJd3N1yfrV9v2r7Hc2bN8887LDDTI/HY3bo0MG87777zCeeeMKUZObl5e3+w9gp7h23+Ph4s0OHDuYJJ5xg/uc//zGrqqoa/JkEAgHzqquuMlu1amUahlHj83nxxRfN7t27mx6Px+zVq5c5efLkBp/jfD6f+cADD5h9+/Y1PR6PmZGRYR544IHm+PHjw8samaZpLl261Bw+fLiZkJBgSmJpP0QFygt7jo3yAuWF5iwvNObx1OeatOO4dtxcLpeZmZlpDh482Lzllltq/W1VVVWZ1113ndm2bVszISHBHDp0qDl79ux6/b7qcr3ftm2becUVV5i9evUyk5KSzLS0NHPw4MHmm2++uUtMX3/9tTlq1CgzLS3NjI+PN7t27WqOHTvWnDt3bnifPZUdYoFhmhEwwwgA21i7dq06d+6shx56SNdff73V4aAJXHvttXr22WdVVlbWbJNkAQDshfIC0LwY4w8A2K3Kysoa9wsKCjR16lQNGzaMpB8AACBKMMYfALBbQ4YM0YgRI9S7d2/l5+frxRdf1Pbt23X77bdbHRoAAADqiMQfALBbxx13nN5++20999xzMgxDAwcO1Isvvqjhw4dbHRoAAADqiDH+AAAAAADYGGP8AQAAAACwMRJ/AAAAAABsjMQfAAAAAAAbI/EHAAAAAMDGSPwBAAAAALAxEn8AAAAAAGyMxB8AAAAAABsj8QcAAAAAwMZI/AEAAAAAsDESfwAAAAAAbIzEHwAAAAAAGyPxBwAAAADAxkj8AQAAAACwMRJ/AAAAAABsjMQfAAAAAAAbI/EHAAAAAMDGSPwBAAAAALAxEn8AAAAAAGyMxB8AAAAAABsj8QcAAAAAwMZI/AEAAAAAsDESfwAAAAAAbIzEHwAAAAB2MmLECF177bVWhwE0GhJ/AAAAABFj7NixMgxjl9sxxxxjdWhNzjAMTZs2zeowYEMuqwMAAAAAgJ0dc8wxmjx5co1tHo/HomiA6EeLPwAAAICI4vF4lJWVVeOWkZGhGTNmyO1267vvvgvv++CDD6p169bKz8+XJC1YsEBHHHGEEhIS1KJFC1166aUqKysL7z927FidcsopGj9+vFq1aqXU1FT9/e9/l8/n2208U6dO1aBBg5SSkqKsrCydffbZ2rJlS/jxGTNmyDAMffnllxo0aJASExN16KGHatmyZTVe5/3339fAgQMVHx+vLl26aPz48QoEApKknJwcSdJf//pXGYYRvg80BhJ/AAAAAFFhx9j78847TyUlJZo3b55uv/12vfDCC2rTpo3Ky8s1atQoZWRkaM6cOXrrrbf0xRdf6Morr6zxOl9++aWWLFmiGTNm6LXXXtO7776r8ePH7/Z9/X6/Jk6cqPnz52vatGlau3atxo4du8t+t912mx555BHNnTtXLpdLF154Yfix7777Tueff76uueYaLV68WM8++6ymTJmie+65R5I0Z84cSdLkyZO1efPm8H2gMRimaZpWBwEAAAAAUnWL/CuvvKL4+Pga22+99Vbdeuut8vl8Gjx4sHr06KGFCxdq6NCheu655yRJzz//vG666SatX79eSUlJkqSPPvpIJ554ojZt2qQ2bdpo7Nix+t///qf169crMTFRkvTMM8/ohhtuUElJiRwOh0aMGKH9999fjz32WK0xzp07VwcddJBKS0uVnJysGTNmaOTIkfriiy905JFHht/3+OOPV2VlpeLj43XUUUfpyCOP1C233BJ+nVdeeUU33nijNm3aJKl6jP97772nU045pTE/UoAx/gAAAAAiy8iRI/X000/X2JaZmSlJcrvdevXVV9W/f39lZ2dr0qRJ4X2WLFmiAQMGhJN+SRo6dKhCoZCWLVumNm3aSJIGDBgQTvolaciQISorK9P69euVnZ29Szw///yz7rrrLs2fP19FRUUKhUKSpHXr1qlPnz7h/fr37x/+f9u2bSVJW7ZsUadOnTR//nzNnDkz3MIvScFgUFVVVaqoqKgRD9DYSPwBAAAARJSkpCR169Ztt4/PmjVLklRYWKjCwsIaiX5j2zF8YNSoUXr11VfVqlUrrVu3TqNGjdplXoC4uLjw/w3DkKRwJUFZWZnGjx+vv/3tb7u8x597NwCNjcQfAAAAQNRYtWqVxo0bp+eff15vvPGGxowZoy+++EIOh0O9e/fWlClTVF5eHq4MmDlzphwOh3r27Bl+jfnz56uyslIJCQmSpB9++EHJycnq2LHjLu+3dOlSFRQU6P777w8/Pnfu3HrHPXDgQC1btmyPFRpxcXEKBoP1fm1gb5jcDwAAAEBE8Xq9ysvLq3Hbtm2bgsGgzj33XI0aNUoXXHCBJk+erN9++02PPPKIJOmcc85RfHy8xowZo4ULF+rrr7/WVVddpfPOOy/czV+SfD6fLrroIi1evFgfffSR7rzzTl155ZVyOHZNjzp16iS3260nn3xSq1ev1gcffKCJEyfW+5juuOMOvfzyyxo/frwWLVqkJUuW6PXXX9e//vWv8D45OTn68ssvlZeXp6KiogZ8ckDtSPwBAAAARJRPPvlEbdu2rXEbNmyY7rnnHuXm5urZZ5+VVD2O/rnnntO//vUvzZ8/X4mJifr0009VWFiogw46SKeddpqOPPJI/fvf/67x+kceeaS6d++u4cOH68wzz9RJJ52ku+66q9ZYWrVqpSlTpuitt95Snz59dP/99+vhhx+u9zGNGjVK06dP12effaaDDjpIhxxyiCZNmlRjToFHHnlEn3/+uTp27KgDDjig3u8B7A6z+gMAAACIGWPHjlVxcbGmTZtmdShAs6HFHwAAAAAAGyPxBwAAAADAxujqDwAAAACAjdHiDwAAAACAjZH4AwAAAABgYyT+AAAAAADYGIk/AAAAAAA2RuIPAAAAAICNkfgDAAAAAGBjJP4AAAAAANgYiT8AAAAAADbmsjoAAPZmmqb8QSkQqv7XHzQVCFX/u2N7MPT7vpJMU+pbOVeSIRlG9b8Op+RyV9+cnj/+73JLTrfk8vy+LwAAsEL1dd1UICj5QzWv/YGgqaBZfY3fca3v4l+uhMB2VV/vHZLTJRlOyemUnHGS06Mqd5p87jS5nYbcLsnBtR5oMBJ/AA1W6TdV7g2pwmeqwm+q0m+q0lf9b4XfVKUvJG+g+iJfH323flHPZxiSO1GKT5Y8KZIn+Y//xydL8SlSQrrkSarn6wIAENsCQVPbf7/W17jG73TNr/SbCtXzYt/V/4tUvHaP+5S1PlgfmcPC910OKSHOUKK7+pYQ5wj/P3Gn7VQQALsi8QewR4Ggqe1VIZVUmSqtCml7VUjbq6q3+YJWR7eDKfnKq2/K3/1ucfFSUqaU1GKnf1tIyS2qWxcAAIhBoZCpUq9Z4xq/vSqk0qrqJN8qfodb2qmsEQhJpd7qWKvtWhBxGFKKx1BqgkNp8Q6lxhtKS3AoNd4hj4sKAcQuEn8AYd6AqYLykArLgyooD6mgIqQyr3UX/Ebnr5KKN1Xf/iwhXUprK6W3rf43ra0Ul9DsIQIA0JSCIVOFFaHq6/3v/5ZUhurdYt8cfHLX+zkhUyqpMlVSFdT6P1UMxLuk9ESHWiQ51TLJoRZJDiV7mPIMsYHEH4hRwZCprWUhbSuzaZJfX5XF1be8JX9s27kyIL2DlN6+egwiAABRwDRNFVeayi/9/VpfHlRJlSkzSi73PsPTqK9XFZDytoeUtz0U3uZxSS2SnGrxe0VAq2SnEuLoGQD7oQQLxIgdiX7e9qDyS4PaWhaZtfsR5c+VAQ5XdfLfIltqkUNFAAAgopimqaLKkPK3h5RfWn299wasjqrhvGr6YXjegLSpJKhNJX/0DkhPMNQmxam2qU61SXUyRAC2QIkVsCnTNLWtLKRN24PK316d6AdJ9PdNKCAV5lbfVnxbXRGQ0aG6IqBll+qKACYUAgA0o9KqkDYUB8MV+5Ez/86+8zagq39jKK40VVwZ0LItARmqHh7QNtWhrJTqioA4J9d6RB8Sf8BG/EFTm0uCWl8c1MbigKqiuJY/KoQCUsHa6tvybyR3ktS6m9SmR3VFgMuaAgsAwL5CZnUPvg3FQW0oCqikyr61+lVq3K7+DWFKKqoIqagipMV5ATkMqW2qUx0znOqQ7lSimzkCEB1I/IEoV+GrvvivL6qu7adV30K+cmnD/Oqbw1U9HKBNj+pbfIrV0QEAopQ/aGpjSVAbioLaWBKI6u779VFlRt6KOyFT2lgS1Mbfhwa0SHKoY4ZTHdNdykikEgCRi8QfiEJVflNrCwNaUxDQ1rLQ3p+A5hcKSFtXVt8WflQ9JKBdP6ldH8mdaHV0AIAIFwxVJ/trtgW0oTg2K/YrzMjvOVc9aWJIv27wK9ljKCfTpS4tXUpPoBIAkYXEH4gSgZCp9UVBrSkIaFNJkIn5ok3Rhurb4k+lVl2l9v2qewI4I681AwBgDdM0lV8a0uqCgNYVBmw1Xr8hKkORn/jvrMxrauFmvxZu9isz0aEuLV3qnOlUAsMBEAFI/IEIZpqm8rb/UQDw07gf/cyQtGVF9c3lkbJ6VVcCtMhhYkAAiFHFFSGt3BbQ2sKAKnzU7EuSKUNVFk3u1xgKK0IqXOfTz+uktmlOdWnhUqcMp1xMDAiLkPgDEajKb2rlNr+WbwmozEsBwLYC3j/mBEjMkDoNlDruz1AAAIgBwVD1sL3lWxi2VyubTJBr6o/lAuMcUueWLvVqHad05gNAMyPxByLI1rKgluUHlFsYiMmxfDGtokha+mX16gBt+0jZg6SM9lZHBQBoZKVVIS3N92vVNrry74nptH5G/8bmD0nLt1RX9rRJcahnmzh1ynDKQY8/NAMSf8BigZCpNQUBLc8PqKCCGv+YFwpIG3+rvqW1lTodKLXfj7kAACCKmWb1RH1L86vn6cHemS77Jf47yy8NKb/Uq4Q4Qz1au9S9lYulAdGkSPwBi3gDppbk+bU030+NP2pXsllaML26J0DOQVLOwZI7weqoAAB1FDKrK/cXbfaruJKufPURctqjq//eVPpNzd/o14JNfmVnOtW3bZwyE51WhwUbIvEHmlmFL6RFm/1asTWgAA38qAt/pbTiW2n17Op5ALocIsWnWh0VAGA3AkFTK7YGtDjPr3Im62uQoCM2Ev8dQqa0piCoNQVBtUtzar+2ccpKpQIAjYfEH2gm2ytDWpjn1+ptAZbiQ8ME/dKaH6XcudUrAXQ9VEpqYXVUAIDfeQOmluZX9+bzBqyOJroFHfbu6r8nOyYDbJXsUP92cWqfTsqGfce3CGhiRRUh/bbJp3WFQZHvo1GEgtL6X6X186W2vaTuw6WU1lZHBQAxyxuoXr99Wb6f3nyNJBBjLf612VoW0pfLvWqR6Fe/9nHqlEHqhobj2wM0kTJvSL9u8GtNQYCEH03ElDYvkTYvlTr0k3qMkBLSrA4KAGKGP1jdwr9oM/P1NLaAwyMKUNUKKkKascKrlkl+DezoZggAGoTEH2hkVX5TCzb5tGwLXfrRXExpw2/SpkXVywB2Gya5E60OCgBsKxQytXxrQL9t8qvKz8W+KfgMN4n/n2wrD+mzpVVqn+bUwI5uZSSyCgDqjsQfaCT+oKnFeX4t3uyXn25+sEIoWD0HwPpfqycA7HyI5KKrJAA0FtM0tbogoPkb/SrzkpU2Jb/B9Wt3NpYEtamkUp1bOLV/B7eSPVQAYO9I/IF9ZJrVtf7zN/hUxUQ+iAQBr7T8G2ntXKnH4dUrARiG1VEBQFTLLw3qp1yfiiqo3W8OPiN2J/erC1PS6oKg1hZWqmdrl/q3d8vj4lqP3SPxB/bB1rKgflrrUwGFAEQiX7m08CNp/Typ77FSRnurIwKAqFPhC+nn9T6tKWAQf3Pyihb/ugiZ0pL8gFYXBDSwo1vdWrpkUNmPWpD4Aw1Q5Tf1y3qfVm6jiR9RoGSzNOs/Uof9pV5HSJ4kqyMCgIgXCplakh/Qbxt9DOGzAIl//XgD0uw1Pq3YEtDgHLdaJDEBIGoi8QfqIWSaWr4loF83+Ji9F9Fnw69S/tLq2f+zB9H9HwB2Y1NJUHNyvSqpYhy/VapMEv+G2FYe0keLqtSjtUsHdHDLTfd//I7EH6ijrWVB/bjWp0K69SOa+aukRZ9UTwC433F0/weAnVT6Tf2U61VuIbX7VqtSnNUhRC1T0rItAeUWBnRgR7e60P0fIvEH9ioYMjVvg19L8vysKgP72J4nzZpcPft/jxGSk8sBgNi2eltAc9Z55WUUX0SoFJP77auqgDRzjU+rCwIa0tnD7P8xjpIesAdbS4Oaucar7XT1gy2Z0urZUv5yacBJUkYHqwMCgGZX4Qvph7U+bSimlT+SVARp8W8sm7eH9L8FlTqwk1s9WvO5xioSf6AWtPIjppQXSLOm0PoPIOas2urXnHXM2xNpTEmVpluid3qj8YekH9b6lFsY1KGd3Uqi9T/mULoD/oRWfsQmWv8BxI4KX0iz1/i0sYSMPyI53UxA20Q2bw/qg4WVGtTJre6taP2PJST+wO9Cpqn5G/xauJlWfsSwHa3/XQ+Veo6QDFoEANjLuqKAZq320sofwUwX4/ubkj9YvfTfusKgDu3iUUIclSyxgBIdIKncG9KnS6q0gKQfkGRKq2ZKs1+WKkusDgYAGkUwZOrHtV7NWEHSH+lMJ4l/c9hYEtT0hZXaTM+XmEDij5i3viig/y2s1NYylukDaihaL333fHX3fwCIYiWVIX20uErLtjBlfzQIOd1WhxAzKv2mvlhWpV83+BQyaf6yM7r6I2YFQ6Z+We/TknwKAcBu+SuluW9IOQdLvY+SHE6rIwKAelm51a+fcn0KUL8fNYIk/s3KlPTbJr/yS4M6rKtHiW7ahu2IvypiUmlVSJ8sriLpB+pq7U/SrMlSeaHVkQBAnfiDpr5bVaVZa0j6o03QEW91CDEpvzSk/y2s1MZiysd2ROKPmLO+KKDpiypVUEEpAKiXks3S9y/Q9R9AxCutCunjxZVaU8DY5WgUcNDibxVvQPpyuVe/rPfJpOu/rZD4I6b8tsmnr1d45accADRMwCvNfVNa+b3VkQBArTaXBPXhokoVV5K0RCsSf+st3Oz/vczM78guSPwREwIhU9+urNKvG/xWhwLYgCkt+1qa964U5DcFIHIszvPri2VVzNof5fwGiX8k2FAc1EeLK1VaRS9ZOyDxh+2Ve6vH868tpBQANKpNi6TZL0mV262OBECMC4ZMzVzt1dx1PpbltQES/8hRUmnqw0WV2rydcnS0I/GHrW0pDerDxVUqZDw/0DRKNkszX5AK11sdCYAYVeEL6dMlVVq1jQnJ7MJreKwOATvxBaUvllVpaT69/KIZiT9sa9U2vz5bWqUqP3X/QJPylks/TpU2zLc6EgAxprgipI8WV2lbORX8duITLf6RxjSln3J9+mGNVyEm/YtKLqsDAJrCgk0+zWM8P9B8QkFp/geSt0zqOtTqaADEgPzSoL5eznh+O/KS+Ees5VsDqvCbGt7NI5fDsDoc1AMt/rAV0zQ1J9dL0g9YZelX0uLPqpsGAKCJ5BYG9MVSkn67qiLxj2gbioP6fGmVvAGu9dGExB+2EQyZ+m6VV0vyGeMHWGrNj9Kv06p7AQBAI1ua79e3K71ilTH7qjJJ/CPd1rKQPllSqXIfw2yiBYk/bMEfNPXVcmbuByLGpoXSnNelgM/qSADYyLz1Pv2Uy8z9dldB4h8VSipNfby4SsWVJP/RgMQfUa/Kb+qzpVXavJ2TDhBRtq2Wfni5evI/ANgHpmlq1hqvFmxmKF8sqDTjrA4BdVThM/XJ4kptLaXxLdKR+COqVfiquxkVMJsvEJlKNkuzX5KqSq2OBECUCpmmvl/t1cqtDOWLFZUhlvOLJr6g9PmyKm3eTvIfyUj8EbXKvdXr9m6vosMfENHKC6TZL0uV262OBECUCYVMfbfSqzUFJBSxwnTGyTSYLT7aBELS18tJ/iMZiT+iUpk3pE+XVqnUS9IPRIWKwupu/5UlVkcCIEqEQqa+WelVbhGJRCwxnbT2RyuS/8hG4o+oU+YN6bMlVSoj6QeiS0XR7y3/JP8A9iwUMjVjpVfri0kgYo3pIvGPZiT/kYvEH1Gl3BfS50urVOYj6QeiUmWx9MNUqYpu/wBqt6OlfwNJf0wKOZnRP9qR/EcmEn9EjQpfdUs/3fuBKFdR9Hvyz4R/AGoyTVPfraKlP5aF6OpvC4GQ9BXJf0Qh8UdU8AZMfb6MpB+wjfJC6cdXJH+l1ZEAiCA/rPUxpj/GBRwk/nYR/L3ln6X+IgOJPyJeIGjqq+VVKqkk6QdspWybNOd1Kci63ACkeet9WsGSfTEv6KCrv50EQtKXy6tUVMHS21Yj8UdEC5mmvl3l1dYyThaALRVtkH55RwrxGwdi2eI8vxZsphIQUoDE33Z8QemLZVUq9XKttxKJPyLa7DU+JvcB7G7LCmnBh1ZHAcAiq7b5NXedz+owECH8Bom/HVX6TX2xtEpVfnrwWoXEHxHrl/U+rdpGlz8gJmz4VVr6ldVRAGhmG4oCmrWGpB9/8BmM8berUm/18N1AkOTfCiT+iEhL8vxaSJc/ILasmimt+cnqKAA0k21lQX2zyiuTHAA78dHib2vbykP6ZqVXIX74zY7EHxEntzCgOXT5A2LT4k+lvKVWRwGgiZX7Qvp6hVdBhvziT3wi8be7jSVBzcmlrN/cSPwRUYoqQpq52mt1GACs9Os0aXu+1VEAaCKBoKmvl3tVyVhf1MIruvrHgmVbAlq+hd69zYnEHxGjyv/7uB9q/4HYFvRLc9+UfBVWRwKgkZmmqe9Xe1XI0l7YjSrFWR0CmslPuT7lbWcS7+ZC4o+IEAqZ+mZllcp91P4DkFRZLP38thSiQADYya8b/VpXxO/ajr5461ldd0ofXXxYC42/cIRWLZq7230DAb/unTxNXS+epPi/TtCAK5/SJz+vkCRVmdVd/Wd98obGndhTlx/VQf997OYaz9+6KVc3nra/Ksu2N90BocmFTOmblSzz11xI/BERfsr1Kb+UHz2AnRTmSos+tToKAI1kTUFACzbRtdeOfvz8bb32+C06+aJbNP6l79Wx2356+JpTtL1wS637v/PMBL3w/td68rLjtfjpK/X34w7SX+95TfNWbVal6VZp8Tb9594rNPrqe3TDE+9r1iev69fvPw4//+UHx+mMK8YrITm1uQ4RTcQbkL5eXiU/M/03ORJ/WG5pvl/Lt7JsH4BarPtZyt19qxGA6LCtLKhZzOFjW5+89m8dfvJYDT/xPLXv0ltjb35C7vgEffu/qbXuP+vj13TjeSfquIN6qEtWpi4/7mAdN6iHHnlvpirNOG3ZuFaJSakafPRp6tLnQPU+cLg2rVkmSZr96ZtyuuI0aOTJzXmIaELFldVDgExm+m9SJP6wVN72IDP4A9izRZ9KBblWRwGggXwBU9+u9IoGPXsK+H1au3Se+h48MrzN4XCo70EjtXJB7Uu0+n0+xXtqjuVPcLv0/eJ1qjA9yurYVd6qSuUum6+ykkKtWfyLOnbfT+Xbi/Tuc3frvBseadJjQvNbXxTU/I30CGpKJP6wTKXf1Hes3wtgb8yQNO9dyVtudSQAGmDmaq/KmMPHtkqLCxQKBpWW2brG9rTM1ioprH2Fln6HHKknXv9EKzYWKBQK6fN5K/Xu7CXaXFiqkBxKSs3QJXc+q+fGX6LxF47Q0OPOUr9DjtLrT9ymo067TFs3rdXt5x2qW886SHO+fK85DhPNYMEmvzaXMAdIU3FZHQBik2ma+n4VS/kAqCNvmTT/femgsyTDsDoaAHW0eLNf64spyKOmc/75oD6951T1uvwJGTLUtW2GLjjqAP3n81/C+wwacZIGjTgpfH/pL99p/cqFOvf6h3Xjqf11+cTJSmvRRuMvGKGeBwxV6p8qHhB9TEnfrfbqxP0SlBDHtb6x0eIPSyzc7Ndmlu8AUB9bV0mrZ1kdBYA62loW1C8bGM5ndynpLeRwOlXyp4n8Sgq3KC2zTa3PSc1opbfuu1blb/9Luf/5p5Y+c7WS493q0rZlrfv7fV699OA4jb35CeWvX61gMKBeAw9T2+weyurUbY8rCCC6VPlNfbeqivH+TYDEH81uS2lQvzKGB0BDLJshFW2wOgoAe+H9fVx/iLK77bni3MrpdYAWz5kR3hYKhbR4zgx163fwHp8b745T+5apCgRDemfWYp0w7IBa9/vgPw+o/5CjldNrf5mhoELBPxqPggG/Qiz9ait520OsANIESPzRrLwBxvUD2Ac7xvv7K62OBMBu7BjOV864/phxzFlX6pv3p+j7D1/VpjVL9dID18hbVaHDTjhXkvTsXZfozafuDO+/auEcTftmjlbnFeq7hWt1zB0vKxQy9c9zTtzltTeuXqIfv3hXf7v0X5Kkttk9ZBiGvvngJf36/SfanLtcnXsf2DwHimYzf6Nf+fQOblSM8UezmrWaggCAfVRZIs3/nzToDKsjAVCLpfkBbWSCrpgy+OjTtL14m9597m6VFOSrU4/+uv6x95TWorqrf2H+ejkcf7Q3+n1VGv/8O1qzcYuSE9w67sDumnrdqUpJy6zxuqZpavJ9V+vsa++TJyFJkuSOT9Aldzyrlx/6pwI+r869/hFltm7XfAeLZmFK+m6VVyfsl6B4xvs3CsNkAAWaydJ8v37KZawf9u78rY9aHQKiQd9jpZxBVkcBYCcllSFNX1SpYMjqSBDpzvC/q/jitTW2lbbYT+85/mJNQIhInTKcGtE93uowbIGu/mgWpVUh/bKepB9AI1r6hVRRZHUUAH4XMk3NXO0l6UeD+R1uq0NAhFlXFNTqbQGrw7AFEn80OdM0NWuNVwEKAgAaU9Bf3eWfjmtARFi02a9t5Vzs0XB+h8fqEBCBfsr1qtLHuWVfkfijyS3bElB+KT9WAE2gMFfKZRknwGpFFSHNZ8Ue7CO/QYs/duULSrPX0nN4X5H4o0nRxR9Ak1v6JV3+AQuFQqa+X83Sfdh3JP7YnQ3FQa3aRuXiviDxR5Ohiz+AZkGXf8BS8zf5VVTBxR77zisSf+zenFyfKujy32Ak/mgydPEH0Gzo8g9YoqgipIWbaYVD46gSY/yxe76gNHsNPYkbisQfTaLMSxd/AM1s6ZdSZYnVUQAxwzRN/bjWS2cbNBqv4qwOARFuY0lQawuZ5b8hSPzRJOau89HFH0DzCvqlxZ9bHQUQM1ZtC2hLGRd7NB5a/FEXc3N98gepcawvEn80uk0lQa0rClodBoBYlLdE2rba6igA2/MGTP1Mzz40skqTFn/sXYXf1G+sIlJvJP5oVMGQqZ9yvVaHASCWLfxEClH5CDSln9f75KW3LRpZpcnkfqibJfl+FVfS46g+SPzRqBbn+bW9iq43ACxUXiCt+cHqKADb2loa1MqtZP1ofJUhEn/UTciUflpLY2N9kPij0ZR7Q1qwiW43ACLAiu+kyu1WRwHYTsg09UMuXfzR+EyHUwE5rQ4DUSSvNKQ1BVRC1hWJPxrN3PVM6AcgQgT90hIm+gMa28qtARVVcLFHE3DFWx0BotDcdUz0V1ck/mgUeduDyi1kTC2ACLJ5sVSw1uooANvwB03NZ0ItNJGQk27+qL9Kv6mFmzkv1QWJPxrFL8zsCyASLf3K6ggA21ic51eln5Y1NA3TyVJ+aJgleX5V+uiJtDck/thnuYUBbSvnxwYgAhVvlPKWWh0FEPUq/aYW0aqGJhSkxR8NFAhJ85lnbK9I/LFPQqapXzfQ2g8ggi37WjKpnAT2xfyNzOODphV00OKPhluxNaDtVZyk9oTEH/tk1baASli+D0AkK9smbfjN6iiAqFVSGdIKlu9DEws6aPFHw5mmNI+hx3tE4o8GC4aY5AdAlFj+jRQkcQEa4pcNPpnU8aOJBWjxxz7KLQpqWxmTje8OiT8abGl+QBU+SgIAokDVdil3jtVRAFFnW1lQ64soSKPp+WnxRyP4mVb/3SLxR4P4gqYWbOKHBSCKrJwp+b1WRwFEld+YMAvNxG+Q+GPf5ZeGtHk7lZW1IfFHgyzL98vHbwpANPFXSuvmWh0FEDUKK4LaUMzFHs3DZ9DVH42Dxsnakfij3oIhU0vyGSsLIAqt+ZGx/kAdLaC1H83IqzirQ4BN5G0PaStj/XdB4o96W7E1oCo/Y/sBRCFvubThV6ujACJeSWVI6wopOKP5+ESLPxoPFZe7IvFHvYRMU4s380MCEMVWzZZM1voF9mTBJr+o4kdzqqLFH41oQ3FQRRVc63dG4o96WVsQVBkz+QOIZpXF0qZFVkcBRKzSqpDWFDAkBs2ryqTFH41r4WbG+u+MxB91ZpomPyAA9rBqpliYHKjdws209qP50eKPxra2IKjSKlr9dyDxR51tKA6quJKiAAAbKN0qbVlhdRRAxKnym1q9jdZ+NL/KEMv5oXGZkhYxRDmMxB91tjiPHw4AG1k92+oIgIizfKtfQer4YYFyk8QfjW91QUC+ACc1icQfdVRUEVJ+KV1lANhI4TqpdIvVUQARIxQytYzlemEB03AqIJfVYcCGAiFpJb2YJJH4o46WbaG1H4AN5f5sdQRAxMgtCqqS5XphBRcT+6HpLMv3y2ReHxJ/7J0/aGoNNWUA7Gjjb1LAa3UUQERYlk8lP6xhuujmj6ZT6jW1eXvQ6jAsR+KPvVq1LSA/vfwB2FHAJ21YYHUUgOWKKkLaUsbFHtYIOWnxR9NayjAmEn/s3XK6+QOws3VzrY4AsNxSWvthoZCDFn80rY3FQZV5Y7tyk8Qfe5S3nSX8ANhc6VapINfqKADL+IOm1hTQGgbrBGnxRxMzJS3bEtvnORJ/7BGt/QBiwjom+UPsWlcUUCC2G8JgsQAt/mgGq7b6FYrhSf5I/LFbvoCpdUVMhAEgBuQtk/xVVkcBWGIVE/jCYkEHLf5oelWB6i7/sYrEH7uVWxhQKHYrxQDEklBA2rzE6iiAZlfuDSl/O839sJbfoMUfzSOWKzpJ/LFbqxnvByCWbFpodQRAs1tdEBB1/LCajxZ/NJMNxUFV+WPzrEfij1qVe0PKL6UFAEAMKciVKrdbHQXQrGK59QuRwy9a/NE8Qqa0tjA2z3sk/qgVs/sCiD0mrf6IKVvLgtpeFZstX4gsPrr6oxnFap5D4o9a0c0fQEzauMDqCIBmQ2s/IoWXFn80o61lIZV5Y69nM4k/dlFUEVJxJS0AAGJQ6RZpe77VUQBNzjRNrYvR7q6IPCT+aG6x2OpP4o9d0NoPIKbR3R8xYEtZSFVc7hEhqkwSfzSvWFyynMQfu1hfREkAQAzLW2Z1BECT41qPSFJpxlkdAmJMQXlIFb7Y6u5P4o8atleGmOgHQGwrL6i+ATYWi61diFyVYjk/NL/1MXYeJPFHDeuLY+sHAAC1yl9udQRAkymsCKrMSyU/IkcFLf6wQKzlPST+qGF9MV3/AIDEH3YWa61ciGym4ZCPxB8WyNselC8YO5WgJP4Iq/Kb2loaW2NdAKBWReslX6XVUQBNgm7+iCguuvnDGiFT2hhDrf4k/gjbWBJQ7NR5AcAemKa0ZYXVUQCNrswbUlEFlfyIHKaTGf1hnVia6JTEH2EbaAEAgD9sobs/7GdTCdd6RJaQkxZ/WGdjSVAhMzaaPkn8IUkKmSaFAQDY2dbVUoiWUdjL5u1c6xFZQrT4w0L+oLStLDau9ST+kFS9lqU/Nr7zAFA3Aa9UstnqKIBGY5qm8kj8EWGCtPjDYnmlsXFeJPGHJFEQAIDaFKy1OgKg0RRVhOSNneGsiBJBBy3+sFas5EEk/pAk5W2nuR8AdlGYa3UEQKPZzLUeESjgoMUf1tpaGlIwZP9x/iT+UChkamtZbNR0AUC9FK5nnD9sg/H9iER+gxZ/WCtoSltjYJw/iT+0rTykgP2/6wBQf0GfVLLJ6iiAfRYKmdoSI+NYEV38tPgjAsRCd38SfyifggAA7B7j/GEDW6nkR4SixR+RgMQfMSEWvugA0GAFjPNH9GNIHyKVTyT+sN628pACQXuP8yfxj3Eh04yJMS0A0GBF6yWT8ySiW6ysU43o4yXxRwQImVJhhb3PkyT+Ma6k0qTrHwDsSdAvlW2zOgpgn2wr52KPyFRF4o8IYffzJIl/jCsop+sfAOxVyWarIwAarNIXUoXP3l1YEb1I/BEp7J4XkfjHuAKb12wBQKMoJvFH9NrKtR4RrMok8UdksHteROIf4+z+BQeARkGLP6JYAeP7EcEqSfwRIbZXmfIF7Ns7isQ/hoVMU0WVFAYAYK+25zHBH6LWNpt3X0V0qyDxRwQpsPEEfyT+Mayk0lTQvt9tAGg8oQAT/CFq0bsPkco0DHnNOKvDAMLsPM6fxD+G2fmLDQCNju7+iELlvpB8XO4RqZweqyMAarBzRSmJfwyz+1qVANCoSvKsjgCot5JK+45XRfQzXXTzR2QptvEwaBL/GFZi4y82ADQ6uvojCnGtRyQzafFHhCmtMhUy7VlhSuIfw7ZX2fNLDQBNorzA6giAeiPxRyQLkvgjwoRMqcxrzxyJxD9GBUKmyn32/FIDQJOo3C4FA1ZHAdRLSRWJPyJXyElXf0Se7TatMCXxj1GltPYDQD2ZtPoj6th5vCqiX8BBiz8iT4lN8yQS/xi1nRYAAKi/8kKrIwDqrMpvyksnFUSwgIMWf0Qeu/aUIvGPUST+ANAAtPgjiti18Ar7CBi0+CPy0NUftsLEfgDQALT4I4qUee1ZeIV9+A1a/BF57NpASuIfo+z6hQaAJkWLP6JIBZP4IsKR+CMSVQWkQNB+508S/xjFjP4A0ACV262OAKgzrvWIdF4Sf0SoCr/9zp8k/jHINE1V2vDLDABNzlsmmZw/ER1o8Uek84nEH5HJjudPEv8YVBWg3AoADWKGJF+F1VEAdUKLPyJdFYk/IhSJP2yh0sf4fgBoMG+p1REAdVLB9R4RrkrM6o/IVOG33/mTxD8G0c0fAPZBVZnVEQB7FQiZ8gasjgLYsyozzuoQgFpV0uIPO7DjZBUA0Gy8JP6IfHbspgr7qTDp6o/IZMd8icQ/BtmxBgsAmk0VXf0R+apsWGiF/VSGSPwRmexYeUriH4Po6g8A+4AWf0QBnw3XoIbdOOjqj4hlx8pTEv8YVBWw3xcZAJoNs/ojCvgY348IF3S4JcOwOgygVn4bVp6S+MegQNDqCAAgigV8VkcA7BUt/oh0AQfd/BG5fDbMl0j8Y5Ada7AAoNkESfwR+Uj8EelI/BHJQqYUDNnrPEriH4NsuCwlADSfgNfqCIC9oqs/Ip3f8FgdArBHfpu1+pP4x6AArQAA0HABv9URAHtFiz8inc+gxR+RzW69pEn8Y5DdvsQA0KyCtPgj8vmZyBcRjsQfkc5uFagk/nVkGIamTZtmdRiNgq7+ALAPbDa534gRI3TttddaHQYaGdd6RDqfSPwR2WK6q//YsWNlGIbuv//+GtunTZsmo57LceTk5Oixxx6r036GYexy+3MMdjRjxgwZhqHi4uJGe82QaSoYxYWB956/R2MGJ9e43XzGAeHHfd4qvfzgOP3j6E66dEQbPXnT2SopyN/ja5qmqXefnairj+uqi4e31ANXnqC8dSvDj/t9Xj1758W6bGRb3Xja/lr009c1nv/R1Mc09eHrGvdAgX1016tfyTjhjhq3Xn9/Ivx4lc+vK56erhZn3afk0+7Wqfe+rvyiPa9Pb5qm7njlS7U970El/G2CjrptilZsLAg/7vUHdN4j7yj19HvU49LH9cWvq2o8/6F3vtdVz3zYuAdqhaBfMhunFWDHdfXPt2OOOaZRXj+S2alCPRKFbDYpFezHK8b4I7IFbHYeddX3CfHx8XrggQd02WWXKSMjoyli2sWECRN0ySWX1NiWkpLSLO9tN9Gc9O/Qvktv3fjv6eH7Tqcz/P//PnaT5s/8VFfe97ISktI09eHr9MTN5+j257/Y7et9NHWSPn/zGV1yx7Nq2S5H7z47UQ9fc4rufX2u3J54zZj2H61dOk+3v/ilfpv1uZ6+4wI9+fEaGYahrZvWasb7UzR+yrdNesxAQ/Tt1Fpf3DMmfN/l+KOud9zzn+jDucv11s1nKi0pXlc+PV1/u/c1zXzoktpeSpL04Dvf64n//aiXxv1Vndtk6PZXvtKoO17W4qevVLw7Ts99Mlc/r9yk2Q9foo9/Xq6zH3pb+a/cKMMwtCavSM9/+rPmPnZZkx5zswn6JVfjtFYdc8wxmjx5co1tHg8FYuwbm5VXYUNVirM6BGCPGqmOP2LUu6v/UUcdpaysLN1333173O+dd95R37595fF4lJOTo0ceeST82IgRI5Sbm6tx48aFWzf2JCUlRVlZWTVuSUlJkqorBdq1a6eCgj9anY4//niNHDlSoVBor7FI1b0KJk6cqLPOOktJSUlq3769nnrqqT3GdNNNN6lHjx5KTExUly5ddPvtt8vv/2PCp7vuukv777+/pk6dqpycHKWlpWn06NEqLS0N7xMKhXTfffepc+fOSkhI0IABA/T2229LktauXauRI0dKkjIyMmQYhsaOHbvHmOrCDgUBp9Ol9BZtwreU9JaSpIqyEn37wcs6+5r71GfQCHXufYAuvv1prfztB61c8FOtr2Wapj59/SmdeMGNGnj4CerUfT9detdzKt62Wb988z9J0qa1y3TA8OPVoUsfHXXapSot2qbS4m2SpJceuFZnXDlBCcmpzXPwQD24nA5lZaSEby3Tqs+bJeVVevHzX/ToRcfoiAFddGC3dpp87V81a8l6/bB0fa2vZZqmHnt/tv515nCdfEhv9e+cpZf/+TdtKizVtNlLJUlL1m/VSYN7qm92a11x/GBtLSnXtu0VkqTL/+9/emDs0UpNjG+eg29qZuPVono8nl2ucRkZGZoxY4bcbre+++678L4PPvigWrdurfz86p5MCxYs0BFHHKGEhAS1aNFCl156qcrK/ui5MXbsWJ1yyikaP368WrVqpdTUVP3973+Xz7f74QpTp07VoEGDwtfes88+W1u2bAk/vqM32pdffqlBgwYpMTFRhx56qJYtW1bjdd5//30NHDhQ8fHx6tKli8aPH69AoHqq+ZycHEnSX//6VxmGEb6PxmOH6z3srcqkqz8im91Oo/VO/J1Op+699149+eST2rBhQ637/PzzzzrjjDM0evRoLViwQHfddZduv/12TZkyRZL07rvvqkOHDpowYYI2b96szZs3N/gAbrvtNuXk5Ojiiy+WJD311FOaNWuWXnrpJTkcjr3GssNDDz2kAQMGaN68ebr55pt1zTXX6PPPP9/t+6akpGjKlClavHixHn/8cT3//POaNGlSjX1WrVqladOmafr06Zo+fbq++eabGkMU7rvvPr388st65plntGjRIo0bN07nnnuuvvnmG3Xs2FHvvPOOJGnZsmXavHmzHn/88QZ/TjvYoeYqb/0qXXN8N13/1/30zB0XqiCvOlFZu3SeggG/+hw8Mrxvu5yeapHVUSsX1p74b920ViUF+eq703MSk9PUpe+gcGVBp279tPzX2fJVVWrBj18ovWWWUtJbatYnbyjOHa9BI05qwqMFGm7FpgK1O/8hdbloks556G2t21IsSfp55Sb5A0EdtX+X8L69OrZSp1Zpmr2bxH9NfpHyisp01P5dw9vSkuI1uGf78HMGdM7S94vXqdLr16e/rFTbzBS1TE3Uq1/PV7zbpb8e2qfpDra5NcPJdMfY+/POO08lJSWaN2+ebr/9dr3wwgtq06aNysvLNWrUKGVkZGjOnDl666239MUXX+jKK6+s8TpffvmllixZohkzZui1117Tu+++q/Hjx+/2ff1+vyZOnKj58+dr2rRpWrt2ba0Vz7fddpseeeQRzZ07Vy6XSxdeeGH4se+++07nn3++rrnmGi1evFjPPvuspkyZonvuuUeSNGfOHEnS5MmTtXnz5vB9NB4Sf0S6ChJ/RDqbnUfr3dVfqq6h33///XXnnXfqxRdf3OXxRx99VEceeaRuv/12SVKPHj20ePFiPfTQQxo7dqwyMzPldDrDrQl7c9NNN+lf//pXjW0ff/yxDjvsMDmdTr3yyivaf//9dfPNN+uJJ57QCy+8oE6dOtUplh2GDh2qm2++ObzPzJkzNWnSJB199NG1xrRzPDk5Obr++uv1+uuv68YbbwxvD4VCmjJlSnhYwnnnnacvv/xS99xzj7xer+6991598cUXGjJkiCSpS5cu+v777/Xss8/q8MMPV2ZmpiSpdevWSk9P3+vnFAu69D1Il9zxjLI69VBJQZ6mvXCf7rnsL7rnvz+ppGCLXHFuJaWk13hOambr3Y7z37E9LbP1rs8prH7ssJPO1/qVC3XL6EFKSW+hK+55WeXbi/Tuc3frlqc/1tvPjNePn7+j1u0766J/Pa3M1u0a/8CBehrcs4OmjPurerZvqc2FpRr/2gwddtOLWvjUlcorKpPb5VR6ckKN57RJT1bebsb579jeJj151+cUVz924dED9dvafPX5x5NqmZqoN286Q0Vllbrj1a80474L9a+pX+j1bxeqa1aG/nPNX9W+ZTT3lGm80sD06dOVnFzzc7311lt166236u6779bnn3+uSy+9VAsXLtSYMWN00knVlY3//e9/VVVVpZdffjncC+7f//63TjzxRD3wwANq06aNJMntdus///mPEhMT1bdvX02YMEE33HCDJk6cKIdj1/r/nRP4Ll266IknntBBBx2ksrKyGnHec889OvzwwyVJN998s44//nhVVVUpPj5e48eP180336wxY8aEX2fixIm68cYbdeedd6pVq1aSpPT09DqVA1B/NiuvwoYqSfwR4ex2Hm1Q4i9JDzzwgI444ghdf/31uzy2ZMkSnXzyyTW2DR06VI899piCwWCNMdl1ccMNN+zS2tC+ffvw/7t06aKHH35Yl112mc4880ydffbZ9Y5lR/K9w5AhQ/Y4+eAbb7yhJ554QqtWrVJZWZkCgYBSU2sWYnNycmrMRdC2bdtwd8mVK1eqoqJil4oFn8+nAw44QE0l2r/AAw79yx93uu+nLn0H6bqT++inL9+V25Ow+yfuA5crTuffWLM3x/MT/q6/nPF35S6br1++ma67X5mtD6dO0quPXK+rHvhvk8QB1Mexg3qE/9+/c5YG9+yg7Asf1ZvfL1SCu2nGVca5nHrq8hNqbLvgsfd09YmHaN7qzZo2e6nmP/kPPfjO97r6uY/0zq2jmySOZtGILf4jR47U008/XWPbjopft9utV199Vf3791d2dnaNnmVLlizRgAEDwkm/VH19C4VCWrZsWTjxHzBggBITE8P7DBkyRGVlZVq/fr2ys7N3iefnn3/WXXfdpfnz56uoqCg8bG7dunXq0+ePXhv9+/cP/79t27aSpC1btqhTp06aP3++Zs6cGW7hl6RgMKiqqipVVFTUiAdAbKoMuaX6zQ0ONKtoz5v+rMHL+Q0fPlyjRo3SLbfc0pjx1Kply5bq1q1bjVtCQs0k79tvv5XT6dTatWvDYwibyuzZs3XOOefouOOO0/Tp0zVv3jzddtttu4yZjIurWbg2DCNcgNoxBvPDDz/Ur7/+Gr4tXrw4PM4fe5eUkq6sTt2Uv3610lq0VsDvU3lpcY19thduUVqLNrU+f8f2ksItNbZvL9yitMzan7Nk7jfauGaJjjr971r6y3cacOgoeRKSdPBRf9OSX77f94MCmkB6coJ6tG+hlZsKlZWRLF8gqOKyyhr75BeXKSsjudbn79ieX1yzR0B+cZmy0mt/zte/rdai3C268oTBmvHbGh03qLuS4t06Y9h+mrFgTSMclT0kJSXtco3bkfhL0qxZsyRJhYWFKiwsbNJYdgwfSE1N1auvvqo5c+bovffek6Q9XuN2zNWz8zVu/PjxNa5vCxYs0IoVKxQfb5N5HiKcaYexfbA1s54rggHNzW6n0QYn/pJ0//3363//+59mz55dY3vv3r01c+bMGttmzpypHj16hFvY3W63gsHGWRzxjTfe0LvvvqsZM2Zo3bp1mjhxYr1ikaQffvihxj4//PCDevfuXev7zZo1S9nZ2brttts0aNAgde/eXbm5ufWKuU+fPvJ4PFq3bt0uBb6OHTtKqv6MJDXa5yTZr2K1qqJMWzauUXrLNsrpdYCcrjgtnjMj/Pjm3OUqyFuvbvsdXOvzW7XLUVqLNjWeU1m2XasXzVW3frs+x+et0ssP/VMX3PyEHE6nQsGgAoHqSR2DgYDMkM0W/IRtlFV6tWpzkdpmpujAbu0U53Lqy/mrw48v27BN67aWaEivjrU+v3ObDGVlJOvLX/94zvaKKv24bGOtz6leLvBDPXvlSXI6HQqGTPl/X1bEHwwqGO0DkI19unzW2apVqzRu3Dg9//zzGjx4sMaMGRNOrnv37q358+ervLw8vP/MmTPlcDjUs2fP8Lb58+ersvKPSp4ffvhBycnJ4WvNzpYuXaqCggLdf//9Ouyww9SrV68aE/vV1cCBA7Vs2bJdrm/dunULDy+Ii4tr1OsbajJsd8UHAOyLfSq59OvXT+ecc46eeOKJGtuvu+46ffnll5o4caKWL1+ul156Sf/+979rDAvIycnRt99+q40bN2rbtm17fJ/S0lLl5eXVuG3fvl2StGHDBl1++eV64IEHNGzYME2ePFn33ntvOJGvSyxSdWHpwQcf1PLly/XUU0/prbfe0jXXXFNrPN27d9e6dev0+uuva9WqVXriiSfCLSJ1lZKSouuvv17jxo3TSy+9pFWrVumXX37Rk08+qZdeekmSlJ2dLcMwNH36dG3durXGTM0NFe2Vq689fquW/vKdtm7K1YrfftATN50lh8OhQ/5yuhKT0zT8pPP12uO3aMncb7RmyTy9MPFydes3uEYSf/MZB2jujA8kVbdSjRp9hT6Y/KB++fZDrV+5UM+Nv1TpLdtq4OEn7vL+H/znfvU/dJSyew6QJHUfcIh+nvGB1q1YqC/eelbd+x/SPB8EsBfXv/iJvlmwRmvzizRryTr99Z7X5HQYOuvwfkpLitdFRw/UP1/4RF//tlo/r9ykCx57T0N6ddQhOyXxvf7+hN6btVhS9W/l2pOH6O43vtEHPy7VgrX5Ov/Rd9UuM0WnDOm1y/tPfP0bHTeouw7oWt0FfGifTnp31mL9tiZP/57+o4b2rr2CIWo04snU6/Xuco3btm2bgsGgzj33XI0aNUoXXHCBJk+erN9++y28Ms0555yj+Ph4jRkzRgsXLtTXX3+tq666Suedd164m79U3VJ/0UUXafHixfroo49055136sorr6x1fH+nTp3kdrv15JNPavXq1frggw9qVKbX1R133KGXX35Z48eP16JFi7RkyRK9/vrru8yP8+WXXyovL09FRUUN+OSwJ7X8eQEAMazBY/x3mDBhgt54440a2wYOHKg333xTd9xxhyZOnKi2bdtqwoQJNcbpT5gwQZdddpm6du0qr9e7xy5pd9xxh+64444a2y677DI9/fTTGjt2rA4++ODwLMajRo3S5ZdfrnPPPVe//vprnWKRqisI5s6dq/Hjxys1NVWPPvqoRo0aVWs8J510ksaNG6crr7xSXq9Xxx9/vG6//Xbddddddf/gJE2cOFGtWrXSfffdp9WrVys9PV0DBw7UrbfeKql6HoMdEyRdcMEFOv/883dZjaC+oj3xL9qyUU/ffoHKSgqVkt5SPQYM0e0vfq3UjOqJos6+9gE5DIeevOVc+X1e9TvkSJ1/42M1XmNz7gpVlm0P3z/uvHHyVpZryn1XqaKsRN0HDNH1j78nt6dmd9QNqxbppy/e08RXZoW3HXTEX7X0l+9072V/UVZ2d10+4T9Nd/BAPWzYtl1nPfS2CrZXqFVakob16aQfHrlUrX5f0m/SJcfI4TB06r1vyOsPaNTAbvq/f9Qcn79swzaVVHjD9288dZjKq3y69MkPVFxepWF9OumTCecp/k9zBixcm683v1uoX5/8R3jbaUP7aMaCNTrsphfVs31L/feG05rw6JtD451MP/nkk/AY+R169uyps88+W7m5uZo+fbqk6nH0zz33nM466yz95S9/0YABA/Tpp5/qmmuu0UEHHaTExESdeuqpevTRR2u81pFHHqnu3btr+PDh8nq9Ouuss3Z7vWrVqpWmTJmiW2+9VU888YQGDhyohx9+ODyhYF2NGjVK06dP14QJE/TAAw8oLi5OvXr1Cq/AI0mPPPKI/vnPf+r5559X+/bttXbt2nq9B/bMReIPAPvEEeV5058ZJoPAlJOTo2uvvVbXXnut1aE0uWDI1KtzK6wOA9ij87c+uvedACsde6vkqN9EtVYYO3asiouLNW3aNKtDQTP7YlmVNpUwlAIAGurIHh61T9/ndvKIQX1wjHE6DNvVXgFAs3I4oyLpR2xzUsIDgH3itFnSxGUhBsVRXgWAhnN5rI4A2CunvcqrANDs7FaBap++C/sg1sYVuhyGvLZbmRIAmonTbXUEdbav88Igerls1lIFAM3NbudRm9VjoC5o8QeAfeCKnsQfsctuLVUA0NzsljNxWYhBcfT/A4CGI/FHFLBbSxUANDe7nUdJ/GMQS/wAwD6Ioq7+iF1uBnMCwD6hxR9RjxZ/ANgHtPgjCsTHca0HgIaKczCrP2zA7bLXlxgAmlVcgtURAHsVz7UeABrMjpWnJP4xKMGGX2QAaDaeZKsjAPbKjoVWAGgudsyXSPxjUKINv8gA0GziU6yOANgrWvwBoOHsWHlK4h+DEtz2+yIDQLOhxR9RwI6FVgBoLrT4wxbs+EUGgGYTT+KPyBfnNMRcvgDQMHasPCXxj0F09QeAfeChqz+igx0LrgDQHOzYUEriH4Ps+EUGgGZDV39EiUSG9gFAg9ix4pTEPwY5HIbiXVZHAQBRyJ0kObh0IjqkePiuAkBDJNmw4pQrQoxKcPOnB4B6Y0Z/RJFkj/0KrgDQHFLi7Zcr2e+IUCepFAYAoP6SMqyOAKizlHiu9QBQX/EuyW3D2VFJ/GNUagJ/egCot6QWVkcA1FkyXf0BoN7s2NovkfjHrFRaAQCg/kj8EUVS6N0HAPVm13MniX+MSrVpTRYANCkSf0SRhDhDTi73AFAvtPjDVkj8AaABkjOtjgCoM8MwlGzDmakBoCnZdUUUex4V9srjMuRhST8AqDt3ohSXYHUUQL2kMacPANSLXSdG5WoQw2j1B4B6oJs/olBGItd6AKiPNJvmSPY8KtQJiT8A1AOJP6IQiT8A1F2yx5DbRYs/bCaTwgAA1F1qG6sjAOqNxB8A6s7O+ZF9jwx71SKJPz8A1FlaW6sjAOot2W0ozml1FAAQHUj8YUsZiQ7ZsyMLADQyw5DSsqyOAqg3wzCUwQR/AFAnJP6wpTinodQEUn8A2KuklpIzzuoogAahuz8A1E2GjXtE2/fIUCctEun/BwB7RTd/RDESfwDYO49LSnLb93xp3yNDnTDOHwDqgMQfUYxrPQDsnZ27+Usk/jGPwgAA1AGJP6JYRqJDcVzuAWCPWibZuyc0l4EYxwR/ALAXTOyHKOcwDLVMpsgHAHvSOsXe50l7Hx32Ks5p2L5bCwDsk9QsJvZD1GudYu+WLADYF4bsf54k44PapPI1AIDdapFjdQTAPmudbO8CLQDsi4xEh+Kc9u4HTcYHZaVSGACA3WqRbXUEwD5rmeyQYe8yLQA0WBubd/OXSPyh6m4tlAUAoBaGIWV2sjoKYJ8xtA8Ads/u3fwlEn9IcjsNZTK7PwDsKrWt5PJYHQXQKFozwR8A1KoNiT9iRVYMfNkBoN7o5g8bYWgfAOwqLd5QfJz9+z+T+EMSE/wBQK2Y2A82kpXqlMP+ZVsAqJdYqRQl24Ok6u4tTPoDADsxHFJmR6ujABpNnNOw/TrVAFBfHdJJ/BFD4pwGY/8AYGcZHRnfD9vpkOayOgQAiBguBy3+iEEd0ikMAEBYmx5WRwA0unYx0rIFAHXRJsUpZ4yMgSLxR1jHDAoDABBG4g8bSk9wKMkdG4VcANibWOnmL5H4Yyep8Q6lxVMYAAAlt5SSMq2OAmgS7WOooAsAexJL50MSf9TQIYPu/gBAaz/srH1a7BR0AWB30hMMJXtiJx2OnSNFncRSdxcA2K3WJP6wr7apTrkoAQKIce1jbH4zTvuooXWyQ57Y+g0AQE3uRCmjg9VRAE3G5TSo6AcQ8zrF2PxmJP6owTAMZvcHENtad5cM5juBveW04FoPIHYlewy1SibxR4zLyYytHwEA1NCur9URAE2ufZpTcZQCAcSonMzYq/zklI9dtE1zKj6O1i4AMciTLLXsbHUUQJNzOgyW8QUQszrHYK8nEn/swmEY6kyrP4BY1K6vZHBpRGyIxRYvAEhPMJSRGHvX+tg7YtRJLNaCAYDa97M6AqDZtE1zyk09P4AYE6uVniT+qFXLZKdS4+nuDyCGJLWQ0tpaHQXQbKq7+8dmARhA7IrVyU1J/LFbXWL0RwEgRtHajxjUtSXXegCxo0WSQ6nxsZkCx+ZRo046UxgAEEva72d1BECza5PiUIqHHn4AYkO3VrGb35D4Y7dSPA61TuYrAiAGZHaSEjOsjgJodoZhqGsMF4QBxA6XI7Z7NJPVYY96tI6zOgQAaHqdBlodAWCZbi1dos0fgN3ltHApzhm7ZzsSf+xRdqZT8bFbMQYgFrgTpbZ9rI4CsEyi26H26UzvD8DeesR47yYSf+yR02GoWyta/QHYWMcDJAdJD2Jbj9axXSAGYG+ZiQ61TI7taz2JP/aqR2u6AAKwK0PKPtDqIADLtU9zKsnN1R6APVG5SeKPOkj20AUQgE217iYlpFkdBWA5wzAoGAOwJZdD6hzDk/rtQOKPOulJYQCAHWUPsjoCIGL0aB0nFyVDADbTJcYn9duB0zvqpF2aU8ms8wvAThIzpFZdrY4CiBgel6GuLanoB2AvvbOYr0wi8UcdGYah3m340QCwkZyDJYMKTWBnfbLimNcHgG10SHcqLYGUVyLxRz10b+WSh4YAAHbgTpQ6HWB1FEDESYl3qGMG8/oAsIc+tPaHkfijzlxOWv0B2ETOwZKT8xlQGwrKAOygRZJDWalUZO5A4o966dkmTnF8awBEM5dbyjnI6iiAiNU6xalWyVzsAUS3/dpSibkzzuqoF4/LUPfW/IgARLFOB0px8VZHAUQ0Wv0BRLPUeEOdGLZUA4k/6q1PlksOZv4BEI0cTqnzYKujACJepwyn0hO42AOITn3bxslgAt8aSPxRb4luB8v9AIhOHQZI8SlWRwFEPMMwNKC92+owAKDekj2GurYgV/kzEn80yH5t41gFC0B0MRxSlyFWRwFEjU4ZTmUkUlQEEF0GtI+Tg+7Ju+BsjgZJiXeoG63+AKJJx/2lpEyrowCihmEY2r89Y/0BRI+0eENdaO2vFYk/GmxA+zg5+QYBiAYOl9R9uNVRAFGnY4ZLLZO42AOIDvt3cDO2fzc4k6PBEt0O9WpDSwCAKND5YMb2Aw00gFZ/AFEgM9HBTP57QOKPfbJf2zi5+X0BiGRx8VLXoVZHAUSt9ukutUqmyAggsu3fgZn894SzOPaJx2Vov7a0BACIYF2HVif/ABpsYEdm+AcQuVolO9QhnbH9e0Lij33WKytOCXHUrgGIQPEpUs5BVkcBRL02KU5l04UWQISicnLvSPyxz1wOg/F/ACJT9+GSk/MT0BgO7OSWk3p+ABEmJ9OpNilUTO4NiT8aRbdWLtb6BRBZUttUL+EHoFEkexzqk0VFGoDI4XJUV0pi78jU0CgchqHB2fzoAESQvsdKBpc5oDHt147hfQAix35t45Tk5lpfF3xKaDStU5zq0oJJNQBEgA79pcyOVkcB2E6c09DADrT6A7BessdQXyYZrzMSfzSqAzu5FccQGwBWcnmkXkdZHQVgW11autQiiSIkAGsN6uiW00EPpLrirI1GlRBnaP/2dPkHYKEeIyRPktVRALZlGIYOznaL4jYAq2SlOtQpk57G9UHij0bXs41LGQl8tQBYILWNlDPI6igA22uV7FSP1hS6ATQ/hyEdnO2xOoyoQ3aGRucwDB2cQ6s/AAv0PYYJ/YBmMrCjW4lu2v0BNK9+7eKUTiNjvfGJoUm0SXGqeytaAgA0o04DpcxOVkcBxIw4p6FDqOgH0IwyEhzqx4R+DULijyYzqJNbSbQEAGgOCWlSbyb0A5pbh3SXcjKZ1RdA0zMkDensloMJ/RqExB9NJs5paEhnxt8AaAb9T6iezR9Aszs42yMPnfwANLE+WXFqmUxFY0OR+KNJtUujyz+AJtZpoNSyi9VRADErPs7QoI50+QfQdFI8hgZ0oIv/viDxR5Ojyz+AJkMXfyAidG0Vp/ZptMQBaBqHdvbIRRf/fULijyZHl38ATYYu/kDEOLSLR/F08rOVyvJSvfrojfrnyb118fCWmnjxkVq9+Ofw4yUF+Xp+wmW65vhuumR4Kz18zSnKW7dyj6+5YfViPXnT2brulD4aMzhZn7721C77zPrkDY07sacuP6qD/vvYzTUe27opVzeetr8qy7Y3zkEi4vVq41KbVCoW9xWJP5oFXf4BNDq6+AMRJSHO0KFdqIizk//ce4UW/vSVLr3red3z6o/ab/ARevDKE1W4ZZNM09TjN56lLRvX6JqH3tCEqTPVIqujHrzqRHkry3f7mr6qSrVq31mn/2O80lq02eXx0uJt+s+9V2j01ffohife16xPXtev338cfvzlB8fpjCvGKyE5tUmOGZElI8GhAxlK1ChI/NFsBnVyKzWeLjoAGkFSC6n30VZHAeBPOqS71LM1Ff124Kuq1Nyv39eZV96tXgcMU5uOXfXXS25T6w5d9NW7zyt//UqtWviTxtz0mLr0OVBts3tozE2Py+et1OzP3trt63bpc6BGX32PDvnL6Ypz71pRtGXjWiUmpWrw0aepS58D1fvA4dq0Zpkkafanb8rpitOgkSc32XEjcjgd0mHdPHLSxb9RkPij2cQ5DR3eLV5OfrsA9oXDJQ08VXLRAgBEogM7uZWWwMU+2gWDAYWCQcV5aibnbk+CVsyfLb/PK0mKc8eHH3M4HIqL82jF/NkNft+sjl3lrapU7rL5Kisp1JrFv6hj9/1Uvr1I7z53t8674ZEGvzaiy0Gd3EpPIF1tLHySaFYZiQ4N6kRhHcA+6PMXKXXX7qEAIoPLYeiwrh7RSBfdEpJS1K3fYH3wnwdUtHWzQsGgZn78ulYu/FHF2/LVNqenWmR11Fv/d6fKtxcp4Pfpw5cfVeGWjSreltfg901KzdAldz6r58ZfovEXjtDQ485Sv0OO0utP3KajTrtMWzet1e3nHapbzzpIc758rxGPGJGkU4ZTPVozi39joi8Wml3PNnHK2x5UblHQ6lAARJu2vaXsA62OAsBeZCY6NbCDW3PX+6wOBfvg0rue14t3X65rT+guh9Op7J7765C/nK61S+fJ5YrTVff/V/+55x/6x9Ed5XA61fegkeo/5C8yZe7T+w4acZIGjTgpfH/pL99p/cqFOvf6h3Xjqf11+cTJSmvRRuMvGKGeBwxVambrfTxSRJJENxODNwUSf1hiSGePCioqVebdtwsDgBiSkC71O8HqKADUUe8sl/JKg9pQTEV/tGrToYtufeZTeSvLVVleqvSWWXrqtvPVul1nSVLn3gdo4iuzVVFWooDfp9SMVhp/4Qh17nVAo8Xg93n10oPjdNldLyh//WoFgwH1GniYJCmrUzetWjRXBxx2XKO9H6xlSBrWxSOPiy5DjY2u/rCE22VoON0AAdSV4age1x8Xv/d9AUQEwzA0rKuHiX1twJOQpPSWWSrfXqSFP3ypA4YfX+PxxOQ0pWa0Ut66lVqz5BcdMLzxKmk/+M8D6j/kaOX02l9mKKhQ8I+KpGDAr1CIiiU72b9DnLJYuq9J0OIPy7RMdmpgR7fmrqMbIIC96H2UlN7O6igA1JPbaWhEt3h9tLhSgZDV0aC+FvzwhUzTVNvs7spfv1pvPHmb2mb30GEnnidJ+unLd5WS3lItsjpqw8pFenXSjTpw+Anqd8iR4dd49q5LlNGqnc64YrwkKeD3aeOapeH/F23dpNzlvyk+IUltOnat8f4bVy/Rj1+8q4lTZ0qS2mb3kGEY+uaDl5SW2Uabc5erc2+Gf9lFdoZT/doxF1hTIfGHpfpkxamwPKTVBQGrQwEQqdr3lzoPtjoKAA2UnujQoZ09+naV1+pQUE8VZSV66//uUtGWjUpKzdCgkSfrtMvvlMtVPela8bY8vfbYLSop3KL0llkaeuxZOvmim2u8RmH+ejkcf3QyLtq6WXecd2j4/sevPq6PX31cvQYO0y1PfxLebpqmJt93tc6+9j55EpIkSe74BF1yx7N6+aF/KuDz6tzrH1FmayqF7SA9wdChXRjX35QM0zQZZA1LBUOmPl1SpW3lNAWg2vlbH7U6BESK9PbSIedLTuqpgWg3d51Xi/Oo6AdQk9spHdc3QanxjEJvSny6sJzTYWhEd48S4hgDCGAnnhTpwNNJ+gGbGNjRrawUip4A/mBIv88FwrmhqfEJIyIkuh0a2d0jJ7k/AElyuKRBZ0jxKVZHAqCROAxDw7vFK8nNxR5AtQEd4tQhnQr+5kDij4jRMtnJmp0AqvU/gcn8ABuKjzN0ZM94uZm0G4h52ZlO9WsbZ3UYMYPEHxGlS0uX+mZxAgBiWpchUvt+VkcBoImkJzg0ons8S/oCMax1ikPDunhkGJwImguJPyLOwI5xys6kKQCISW17S72O3Pt+AKJaVqpTQzqzbBcQi9LiDY3sHi8ntX/NisQfEccwDA3r4mECICDWtMiR9v+rRO0/EBO6tozTgPb08gNiScLvw308Lq71zY3MChHJ6TA0oke8MhL5igIxITVLOvAMyUFvHyCWDGjvVteWTOwFxAKXQzqih0fJHsr3VuBTR8RyOw0d1cOjZA81goCtJWZIB58lxTG5JxCLhuS41TaVIilgZ4YhHd7NoxZJVPBbhbMsIlqC26GjesYrnsYAwJ7cSdLBZ0ueZKsjAWARh8PQiO7xapVMsRSwqyE5brVn2T5LcYZFxEuNd+jInvFy8W0F7MXlrm7pT8q0OhIAFotzGjqSIX6ALR2c7Va3VsznYTXOrogKLZKcGtkjXk6+sYA9OFzSgWdKaW2tjgRAhHC7DB3VM16p8QzxA+xiUCe3erUh6Y8EpFGIGm1TndVLf1AeAKKbwyUNOlNqmWN1JAAiTEKcob/0ilcK8/sAUW9ghzj1ySLpjxQk/ogq7dKcGtHdI5b9BKKUwykdeLrUqovVkQCIUIluh47uFa9kNxd7IFoNaB+n/dq5rQ4DOyHxR9Rpn+4i+Qei0Y6kv3U3qyMBEOGSPQ4d3TteSST/QNTp1zZOA9qT9EcaEn9EpQ7pLo3s7qHbPxAtdnTvb93d6kgARIkUj0OjetPtH4gmfdvG6YCOJP2RiMQfUat9uosJ/4Bo4IyTDhottepqdSQAokyyx6Fj+iQoI4GLPRDp9m8fpwNJ+iMWZ1FEtXZpTh3VM15up9WRAKiVK146+GypZWerIwEQpRLiDP2ld7xaJlFsBSLVoE5u9ad7f0TjDIqo1ybFqVG9E5QQR1dAIKJ4UqQh50uZnayOBECU87gMHd0rXlkpFF2BSGJIGtLZzez9UYCzJ2whI9GhY/uw9i8QMZJaSIeOlVLbWB0JAJuIcxo6sme8OqTTzQ+IBE5DOry7R91bkfRHAxJ/2Eayx6FjeifQFRCwWnq76qQ/Md3qSADYjNNhaEQ3j7q0cFkdChDT4pzSkT3j1SmD32K0IEOCrcTHVXcFbJdGawBgiVZdpcHnSe5EqyMBYFMOh6FhXT0a0J5WRsAKSW5Dx/ROUFYq5e1oQuIP24lzGjqiO60BQLNr3696yT4Xk/sAaHoD2rs1rItHDkb5Ac2mZZJDx/WJV0YiaWS0ITOCLe1oDUhLMPTrBr9MqwMCbM2Qeo6Uug21OhAAMaZLS5eSPIZmrKiSN2B1NIC9ZWc6NayLR05q26ISVTWwtX7t3BrR3aM4vulA03C5pUFnkPQDsEybFKeO7ZPABL9AE+rXLk7Du5L0RzPSIdhexwyXju2boBQPJyqgUSVmSodeKLXpYXUkAGJcarxDx/ZJUBuW+wMalcOQhnZx64AObhkGZeloxtkRMSE9waHj+iYoK5WvPNAoWnaWhl0opbSyOhIAkCR5XNUT/LKeONA4EuIM/aVXvLq25DdlB2RBiBkel6GjesarVxumtgD2Sc7B0sFnS3EJVkcCADU4DEODOrl1eDeP4phwHGiwrFSHTtgvQa1T+CHZBRkQYorDMHRwtkctk5z6Ya1XgZDVEQFRxOWW9juuevZ+AIhg2ZkuZSQ4NGNllYormeIXqI9+7eI0oH2cHHTttxUSf8SkLi1dapns0LcrvSqsIPsH9io1Sxr4NymphdWRAECdpCZUj/ufvcartYVBq8MBIp7bKQ3r6lGHdFJEO6KrP2JW9URA8erZmpMbsEc5B0mHXkDSDyDqxDkNDe8Wr4Oz3WIycmD3WiRVd+0n6bcv/rKIaU6HocE5HrVNdWrWGq98NAgAf4iLl/qfKGX1sjoSANgnvdrEqU2KU9+tous/8Ge927g0sKObpfpsjsQfkNQp06XMpOqu/9vK6foPKL2DdMBfpcR0qyMBgEaRkejQ8X0T9Mt6n5bkB6wOB7BcktvQ0C4eZaUygV8sMEzTpNoT+F3INLVws1+/bfQrxC/DMudvfdTqEGKXwyl1O0zqOlRyMBoMgD1tLglq5hqvKnxc7BGburRw6eBst9wuWvljBYk/UIuiipBmrfaqgIn/LEHib5G0ttKAk6SU1lZHAgBNzhsw9eNaJv5DbPG4pENyPMrOpON3rCHxB3aD1n/rkPg3M1r5AcSw1dsCmrPOKy+9/2Fz7dOcOrSzWwlurvWxiKoeYDcchqH+7dzqmO6i9R/2RSs/gBjXpaVL7dKcmrvOp9UFZP+wn3iXNKiTR11akvrFMlr8gToImaYWbfbrt01+Bcn/mxwt/s3A4ZK6HyZ1OZRWfgD43eaSoH5Y61Wpl+Ix7KF7q+oZ+z2M5Y95JP5APZR5Q5qT69P6YsYDNiUS/ybWpofUZxQz9gNALYIhU79t9GtRHkP9EL3SEwwdkuNR6xRm7Ec1En+gATYWB/RTro8WgSZC4t9EEjOlvqOk1t2sjgQAIl5RRUg/rPVqaxld/RA9nA6pf7s49c2Kk8NBKz/+QOIPNFAwZGpRnl8LN/kVoEzQqEj8G5kzrnrivi5DJCfj+wCgrkzT1JqCoOZt8Kmcpf8Q4TplOHVgR7dS4hnCh11RAgQayOmonvyvSwuX5q7zaV0R3f8RgbJ6SX3+IiWkWR0JAEQdwzDUpaVL2ZlOLc7zawGV/YhALZMcOrCTW23o1o89oMUfaCRbSoP6Zb1PW+gSuM9o8W8EmZ2knkdImR2tjgQAbKPSF9K8jX6t2hoQBWhYLdlt6ICObuVkOmUYdOvHnpH4A41sQ1FAv2zwqbiSn1ZDkfjvg5TW1Ql/m+5WRwIAtlVYEdTcdT7lbaeyH80vzin1axen3m3i5GQcP+qIxB9oAqZpanVBQPM3+FXGmMB6I/FvgIR0qcfhUvt+ErX+ANAs8rYHNX+jT/mlVACg6bkcUo/WLu3X1q34OK71qB/G+ANNwDAMdW0Zp5xMl5ZvCWjhZr8q/VQAoAl4kqsn7ss+UHIwtg8AmlNWqlNZqQnavD2o+RsY7oem4XJIPdtUz9RPwo+GosUfaAbBkKmV2wJatNmvMpYA3Cta/OsgMaN6lv4OA5ipHwAixOaSoH7d6GMJQDSKuN8T/j4k/GgElBaBZuB0GOrZOk7dW7m0tiCohZuZAwANlNK6uoW/XR/JYLkeAIgkbdOcapuWoE0lQS3YxBAANEycU+r1e8LvcZHwo3GQ+APNyPH7skCdWzi1sTioBZv9tAqgbjI6Vif8TNoHABGvXZpT7dISVFAe1KLNfuUWBUUfW+xNssdQrzZx6tbSJTcJPxoZXf0Bi+WXBrUs3691RUGF+DVKoqt/mOGQsnpJOQdVL88HAIhKZd6QluYHtHKrX76g1dEg0rRJcah3Vpw6prMsH5oOiT8QISr9plZs8Wv51oAqYnwlgJhP/ONTpU4DpU4HVE/eBwCwBX/Q1OptAS3N96ukKrav9bHOYUidW7jUO8ulzEQm50XTI/EHIkzINLWhuLoXwOYYXR84ZhP/ll2qZ+dv04Px+wBgc/mlQa3cGlBuYUCB2Lzcx6QUj6FurVzq1ipOCUzYh2ZE4g9EsO1VIa3YGtCagtjqBRBTiX98qtR+P6nj/lJSC6ujAQA0M3/Q1NqCgFZsDWhbOTUAduRySJ0yXOreyqU2qbTuwxok/kAUME1T+aUhrS6obhnw23x8oO0Tf5dHattbat9PysyWGM8HAJBUXBnSyq1+rS4IqspPET3atUlxqGtLl7IzXYpzcq2HtUj8gSgTDFUPBVhdENDGYntOCGjLxN/hlFp1q072W3eXnCyqAgCo3Y4K/9zCgNYVBVVJJUDUaJXsUKcMl7IznUr2MGwPkYPEH4hi3oCpDcUBrS8KalNJ0DZjBG2T+DvdUqsu1WP22/SQ4hKsjggAEGVM09SW0pDWUgkQkQxVt+x3ynSpU4ZTiW6SfUQmEn/AJoIhU3nbg1pfHNSG4mBUzwkQ1Yl/fOofiX5mNi37AIBGY5qmtpSFfq/wD6i4Mnqv9dHM6ZCyUpzqlOFUxwyX4pmkD1GAEilgE06HofbpLrVPr/5ZF5RXVwBsLglqW3nIlkMCIoLDKaW1q27Zb91dSmtrdUQAAJsyDENtUpxqk+KU5FaFL6RNJcHwzWfzOYCslJnoUNs0p9qlOtU6xSGng2Qf0YUWfyAGBILVLQT5pUHlbQ+qIMIrAiK6xd9wSOntpRbZ1beMjpIzzuqoAAAxzjRNFZRXVwTkba+u9LfLEEArJLkNtU11qm2aU21TnbTqI+qR+AMxKFwRsD2oreVBFZaHIqqVIKISf1e8lJb1e7KfI2WS6AMAIl/INFVUEdLWspC2lgW1tTSksigeBtiUnA6pRaJDLZOdapXsUKtkB2P1YTsk/gAkSaVVIRWUh1RQEVKBxZUBliX+O5L8tLZ/3JIyrYkFAIBGVukLaUtZSIXlIRVVhlRcEXuVAQ5DSktwKCPBoZbJDrVMcigz0SEHXfdhc4zxByBJSol3KCXeoZwWf2wrraouGGyvCml7lRn+N+rXFvYkSUktfr9lVv+b0ookHwBgawluh7IzHcre6XLnD5oqrqi+3hdVhFRcGVJJZUhVAevibAwOQ0qNN5Se4KhxS4435DBI8hF7SPwB7NaOyoA/8wVNba+srgQo84ZU4TdV6TdV6TNV4a+uGLBsDgHDIXmSpfiUmv8mpv+R7Md5LAoOAIDIEuc01CrFqVYpzhrbAyFTZV5T5d6QyrymynzV1/xyr6lynylvwMJrvSSXo3ocfpLHoSS3oWSPoSS3Q0keQ0luQ4luEnxgZ3T1B9DoTNOUN6DqygB/SP5gdYuCP1hdkNhxPxCq/jdkSjvORKako0relWRIOy7YDqfkcktOT/W/LrfkdEsuj+SKq/7Xk1x9cyf+8TwAANBkfEFTXn91JUD1TaoKmPIFTAVC1df5YKh6yeGQqZ1upgwZchjV4+sNo7qF3mEYv/8ruZyGPC7J4zTkdhny/H5zuyS301Cck2s9UB8k/gAAAAAA2BjTVQIAAAAAYGMk/gAAAAAA2BiJPwAAAAAANkbiDwAAAACAjZH4AwAAAABgYyT+AAAAAADYGIk/AAAAAAA2RuIPAAAAAICNkfgDAAAAAGBjJP4AAACADRmGoWnTplkdBoAIQOIPAAAA2xo7dqwMw9D9999fY/u0adNkGEa9XisnJ0ePPfZYnfYzDGOX259jsKMZM2bIMAwVFxdbHQqAnZD4AwAAwNbi4+P1wAMPqKioqNnec8KECdq8eXON21VXXdVs7w8AOyPxBwAAgK0dddRRysrK0n333bfH/d555x317dtXHo9HOTk5euSRR8KPjRgxQrm5uRo3bly4BX9PUlJSlJWVVeOWlJQkqbpSoF27diooKAjvf/zxx2vkyJEKhUJ7jUWq7lUwceJEnXXWWUpKSlL79u311FNP7TGmm266ST169FBiYqK6dOmi22+/XX6/P/z4XXfdpf33319Tp05VTk6O0tLSNHr0aJWWlob3CYVCuu+++9S5c2clJCRowIABevvttyVJa9eu1ciRIyVJGRkZMgxDY8eO3WNMAJoHiT8AAABszel06t5779WTTz6pDRs21LrPzz//rDPOOEOjR4/WggULdNddd+n222/XlClTJEnvvvuuOnToUKMlv6Fuu+025eTk6OKLL5YkPfXUU5o1a5ZeeuklORyOvcayw0MPPaQBAwZo3rx5uvnmm3XNNdfo888/3+37pqSkaMqUKVq8eLEef/xxPf/885o0aVKNfVatWqVp06Zp+vTpmj59ur755psaQxTuu+8+vfzyy3rmmWe0aNEijRs3Tueee66++eYbdezYUe+8844kadmyZdq8ebMef/zxBn9OABqRCQAAANjUmDFjzJNPPtk0TdM85JBDzAsvvNA0TdN87733zJ2LwmeffbZ59NFH13juDTfcYPbp0yd8Pzs725w0adJe3zM7O9t0u91mUlJSjdu3334b3mfVqlVmSkqKedNNN5kJCQnmq6++Wu9YjjnmmBr7nHnmmeaxxx4bvi/JfO+993Yb50MPPWQeeOCB4ft33nmnmZiYaG7fvr3G+w4ePNg0TdOsqqoyExMTzVmzZtV4nYsuusg866yzTNM0za+//tqUZBYVFe32fQE0P1r8AQAAEBMeeOABvfTSS1qyZMkujy1ZskRDhw6tsW3o0KFasWKFgsFgvd/rhhtu0K+//lrjNmjQoPDjXbp00cMPP6wHHnhAJ510ks4+++x6xzJkyJAa+wwZMqTWY9vhjTfe0NChQ5WVlaXk5GT961//0rp162rsk5OTo5SUlPD9tm3basuWLZKklStXqqKiQkcffbSSk5PDt5dfflmrVq2qx6cDoLm5rA4AAAAAaA7Dhw/XqFGjdMsttzT52POWLVuqW7due9zn22+/ldPp1Nq1axUIBORyNV3RfPbs2TrnnHM0fvx4jRo1SmlpaXr99dd3mTsgLi6uxn3DMMLzDpSVlUmSPvzwQ7Vv377Gfh6Pp8liB7DvSPwBAAAQM+6//37tv//+6tmzZ43tvXv31syZM2tsmzlzpnr06CGn0ylJcrvdDWr9r80bb7yhd999VzNmzNAZZ5yhiRMnavz48XWORZJ++OGHGvv88MMP6t27d63vN2vWLGVnZ+u2224Lb8vNza1XzH369JHH49G6det0+OGH17qP2+2WpEb7nAA0DhJ/AAAAxIx+/frpnHPO0RNPPFFj+3XXXaeDDjpIEydO1JlnnqnZs2fr3//+t/7v//4vvE9OTo6+/fZbjR49Wh6PRy1bttzt+5SWliovL6/GtsTERKWmpmrDhg26/PLL9cADD2jYsGGaPHmyTjjhBB177LE65JBD6hSLVF0Z8OCDD+qUU07R559/rrfeeksffvhhrfF0795d69at0+uvv66DDjpIH374od577716fXYpKSm6/vrrNW7cOIVCIQ0bNkwlJSWaOXOmUlNTNWbMGGVnZ8swDE2fPl3HHXecEhISlJycXK/3AdAErJ5kAAAAAGgqO0/ut8OaNWtMt9tt/rko/Pbbb5t9+vQx4+LizE6dOpkPPfRQjcdnz55t9u/f3/R4PLs8d2fZ2dmmpF1ul112mRkKhcwjjzzSHDVqlBkKhcLPueqqq8yuXbuapaWldYolOzvbHD9+vHn66aebiYmJZlZWlvn444/X2Ed/mtzvhhtuMFu0aGEmJyebZ555pjlp0iQzLS0t/Pidd95pDhgwoMZrTJo0yczOzg7fD4VC5mOPPWb27NnTjIuLM1u1amWOGjXK/Oabb8L7TJgwwczKyjINwzDHjBmz288JQPMxTNM0rat2AAAAAFBfOTk5uvbaa3XttddaHQqAKMCs/gAAAAAA2BiJPwAAAAAANkZXfwAAAAAAbIwWfwAAAAAAbIzEHwAAAAAAGyPxBwAAAADAxkj8AQAAAACwMRJ/AAAAAABsjMQfAAAAAAAbI/EHAAAAAMDGSPwBAAAAALAxEn8AAAAAAGyMxB8AAAAAABsj8QcAAAAAwMZI/AEAAAAAsDESfwAAAAAAbIzEHwAAAAAAGyPxBwAAAADAxkj8AQAAAACwMRJ/AAAAAABsjMQfAAAAAAAbI/EHAAAAAMDGSPwBAAAAALAxEn8AAAAAAGyMxB8AAAAAABsj8QcAAAAAwMZI/AEAAAAAsDESfwAAAAAAbIzEHwAAAAAAGyPxBwAAAADAxv4fbwHTEEJESZIAAAAASUVORK5CYII=\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "> **Reminder:**\n",
        "\n",
        "The goal of machine learning models is to generalize well to unseen data. By applying SMOTE only to the training dataset, the model learns to better represent the **underlying patterns** in the data while avoiding overfitting. This ensures that the model's performance on real-world data is more reliable."
      ],
      "metadata": {
        "id": "8xqKf4B3BdBD"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "\n",
        "\n",
        "---\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "\n"
      ],
      "metadata": {
        "id": "9gsT8myTrVD-"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### DataLoader"
      ],
      "metadata": {
        "id": "isaSCpFDsHba"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Creating a DataLoader allows us, among other things, to load data in batches, making it easier to process large datasets without loading everything into memory at once.\n"
      ],
      "metadata": {
        "id": "8vs8UqJFsWiX"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "It also can shuffle the data, which is important for training models to prevent the model from learning the order of the data!"
      ],
      "metadata": {
        "id": "UDGp2Qs8acFQ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# number of samples in each batch during training\n",
        "batch_size = 64\n",
        "\n",
        "spike_grad = surrogate.fast_sigmoid(slope=25)\n",
        "\n",
        "# initialize a beta value of 0.5\n",
        "beta = 0.5\n",
        "\n",
        "\n",
        "# create a dataloader for the testset\n",
        "test_dataloader = DataLoader(test_dataset, batch_size=batch_size, shuffle=False, drop_last=True)\n",
        "\n",
        "# create a dataloader for the trainset\n",
        "train_dataloader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True, drop_last=True)"
      ],
      "metadata": {
        "id": "n_4T14rxgv2x"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "print(train_dataset.data.head(1))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cCxWV3d1gmbe",
        "outputId": "23870156-9c33-4c48-abbd-80e7a099a2dc"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "   LABEL  FLUX.1  FLUX.2  FLUX.3  FLUX.4  FLUX.5  FLUX.6  FLUX.7  FLUX.8  \\\n",
            "0      2   93.85   83.81    20.1  -26.98  -39.56 -124.71 -135.18  -96.27   \n",
            "\n",
            "   FLUX.9  ...  FLUX.3188  FLUX.3189  FLUX.3190  FLUX.3191  FLUX.3192  \\\n",
            "0  -79.89  ...     -78.07    -102.15    -102.15      25.13      48.57   \n",
            "\n",
            "   FLUX.3193  FLUX.3194  FLUX.3195  FLUX.3196  FLUX.3197  \n",
            "0      92.54      39.32      61.42       5.08     -39.54  \n",
            "\n",
            "[1 rows x 3198 columns]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "L2wgL_qcMHBo"
      },
      "outputs": [],
      "source": [
        "fig = make_subplots(rows=2, cols=2,subplot_titles=(\"Star #0 (Exoplanet)\", \"Star #1 (Exoplanet)\",\n",
        "                                                   \"Star #3000 (No-Exoplanet)\", \"Star #3001 (No-Exoplanet)\"))\n",
        "fig.add_trace(\n",
        "    go.Scatter(y=train_dataset.__getitem__(0)['feature']),\n",
        "    row=1, col=1\n",
        ")\n",
        "fig.add_trace(\n",
        "    go.Scatter(y=train_dataset.__getitem__(1)['feature']),\n",
        "    row=1, col=2\n",
        ")\n",
        "fig.add_trace(\n",
        "    go.Scatter(y=train_dataset.__getitem__(3000)['feature']),\n",
        "    row=2, col=1\n",
        ")\n",
        "fig.add_trace(\n",
        "    go.Scatter(y=train_dataset.__getitem__(3001)['feature']),\n",
        "    row=2, col=2\n",
        ")\n",
        "for i in range(1, 5):\n",
        "    fig.update_xaxes(title_text=\"Time\", row=(i-1)//2 + 1, col=(i-1)%2 + 1)\n",
        "    fig.update_yaxes(title_text=\"Flux\", row=(i-1)//2 + 1, col=(i-1)%2 + 1)\n",
        "\n",
        "fig.update_layout(height=600, width=800, title_text=\"Exoplanets Flux vs No-Exoplanet Flux\",showlegend=False)\n",
        "fig.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "---\n"
      ],
      "metadata": {
        "id": "9StLNF7yluaT"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Model network"
      ],
      "metadata": {
        "id": "aHnkDQ6ZmIP4"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "- Let's define a neural network model called `Net` using PyTorch's `nn.Module`."
      ],
      "metadata": {
        "id": "tYv4bxbfmSJ3"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "juQe2MaMwiC-"
      },
      "outputs": [],
      "source": [
        "class Net(nn.Module):\n",
        "    def __init__(self):\n",
        "        super().__init__()\n",
        "\n",
        "        # Initialize layers (3 linear layers and 3 leaky layers)\n",
        "        self.fc1 = nn.Linear(3197, 128) # takes an input of 3197 and outputs 128\n",
        "        self.lif1 = snn.Leaky(beta=beta, spike_grad=spike_grad)\n",
        "        self.fc2 = nn.Linear(64, 64) # takes an input of 64 and outputs 68\n",
        "        self.lif2 = snn.Leaky(beta=beta, spike_grad=spike_grad)\n",
        "        self.fc3 = nn.Linear(32, 2) # takes in 32 inputs and outputs our two outputs (planet with/without an exoplanet)\n",
        "        self.lif3 = snn.Leaky(beta=beta, spike_grad=spike_grad)\n",
        "        self.softmax = nn.Softmax(dim=1) # softmax applied with a dimension of 1\n",
        "\n",
        "    def forward(self, x):\n",
        "\n",
        "        # Initialize hidden states and outputs at t=0\n",
        "        mem1 = self.lif1.init_leaky()\n",
        "        mem2 = self.lif2.init_leaky()\n",
        "\n",
        "        cur1 = F.max_pool1d(self.fc1(x), 2)\n",
        "        spk1, mem1 = self.lif1(cur1, mem1)\n",
        "\n",
        "        cur2 = F.max_pool1d(self.fc2(spk1), 2)\n",
        "        spk2, mem2 = self.lif2(cur2, mem2)\n",
        "\n",
        "        cur3 = self.fc3(spk2.view(batch_size, -1))\n",
        "\n",
        "        # return cur3\n",
        "        return cur3\n",
        "\n",
        "# device = torch.device(\"cuda\") if torch.cuda.is_available() else torch.device(\"mps\") if torch.backends.mps.is_available() else torch.device(\"cpu\")\n",
        "model = Net() # initialize the model to the new class."
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "This neural network consists of three fully connected linear layers (`fc1`, `fc2`, and `fc3`) followed by three leaky integrate-and-fire (LIF) layers (`lif1`, `lif2`, and `lif3`).\n",
        "\n"
      ],
      "metadata": {
        "id": "8e4rKyxiDhix"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "The sizes of the input and output dimensions for each linear layer are specified and a softmax layer (`softmax`) is defined to normalize the output probabilities."
      ],
      "metadata": {
        "id": "hh6VLnhzSEHn"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "For each linear layer, the program specify the number of input and output features. For example, the first linear layer takes 3197 input features and produces 128 output features."
      ],
      "metadata": {
        "id": "DDGtQ1BRMO33"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Then, the program initialize the hidden states and outputs for each neuron in the network at time step 𝘵 = 𝟶. This is like setting up the initial conditions before we start processing the input data."
      ],
      "metadata": {
        "id": "e5okAG9iSJU9"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Following it's the processing of the input data through each layer of the network.\n",
        "\n",
        "For the first linear layer (`fc1`), the program passes the input data (`x`) through the layer and apply a max pooling operation. Max pooling helps reduce the size of the data while retaining important features.\n",
        "\n",
        "Next, it applies the leaky integrate-and-fire (`LIF`) neuron model to the output of the first layer. This simulates how neurons in the brain fire in response to stimuli.\n",
        "\n",
        "This process is repeated for the second linear layer (`fc2`) and its corresponding LIF neuron, and then for the third linear layer (`fc3`).\n",
        "\n",
        "\n",
        "Finally, we obtain the output of the last linear layer (`fc3`) and return it as the result of the forward pass."
      ],
      "metadata": {
        "id": "E8nb0xFyO-vX"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "> **Key takeways:**\n",
        "\n",
        "- The data flows sequentially through each layer of the neural network.\n",
        "\n",
        "- However, within each layer, there is *parallel processing* happening. Different parts of the data are being processed simultaneously."
      ],
      "metadata": {
        "id": "VGtSCy_gQ2zo"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "---"
      ],
      "metadata": {
        "id": "f3auu6tsS4m3"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Loss function and optimizer"
      ],
      "metadata": {
        "id": "hYOi8Y-DS7Wg"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Originally, this model didn't use binary cross-entropy loss, but due to the nature of our classes, it can be well-suited for calculating the loss between the predicted probabilities and the true binary labels.\n",
        "\n",
        "The optimizer continues to be the same. It updates the parameters using a small batch of training examples at a time, rather than the entire dataset (hence \"stochastic\")."
      ],
      "metadata": {
        "id": "CTDJKUBOTNkj"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "criterion = nn.BCEWithLogitsLoss()  # Binary Cross-Entropy Loss\n",
        "optimizer = optim.SGD(model.parameters(), lr=0.001) # stochastic gradient descent with a learning rate of 0.001"
      ],
      "metadata": {
        "id": "Rw5ozGRLV6MM"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "---"
      ],
      "metadata": {
        "id": "jxc_p6wK3fkI"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Training and testing"
      ],
      "metadata": {
        "id": "tCCxucgg3hGn"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Following this model original developer's advice, I left the number of epochs between 50 and 500."
      ],
      "metadata": {
        "id": "RHQWXcDv3moe"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "DXoxzKw9kNJ-"
      },
      "outputs": [],
      "source": [
        "# create a pandas dataframe to hold the current epoch, the accuracy， sensitivity, specificity, auc-roc and loss\n",
        "results = pd.DataFrame(columns=['Epoch', 'Accuracy', 'Sensitivity', 'Specificity', 'AUC-ROC', 'Test Loss'])"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "2d_9gWYbkTGp"
      },
      "outputs": [],
      "source": [
        "num_epochs = 100 # initialize a certain number of epoch iterations"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "PdR9HaAMybL7"
      },
      "outputs": [],
      "source": [
        "\n",
        "for epoch in range(num_epochs):\n",
        "    model.train()  # Set model to training mode\n",
        "    for data in train_dataloader:\n",
        "        inputs, labels = data['feature'].float(), data['label'].float()  # Float\n",
        "        optimizer.zero_grad()  # Clear previously stored gradients\n",
        "        outputs = model(inputs)\n",
        "        loss = criterion(outputs, labels)  # Compute binary cross-entropy loss\n",
        "        loss.backward()  # Backpropagation\n",
        "        optimizer.step()  # Update parameters\n",
        "\n",
        "    # Test Set Evaluation\n",
        "    model.eval()  # Set model to evaluation mode\n",
        "    with torch.no_grad():\n",
        "        test_loss = 0.0\n",
        "        correct = 0\n",
        "        total = 0\n",
        "        all_labels = []\n",
        "        all_predicted = []\n",
        "        all_probs = []\n",
        "        for data in test_dataloader:\n",
        "            inputs, labels = data['feature'].float(), data['label'].float()\n",
        "            outputs = model(inputs)\n",
        "            test_loss += criterion(outputs.squeeze(), labels).item()  # Compute test loss\n",
        "            predicted = torch.round(torch.sigmoid(outputs))  # Round to get predicted labels\n",
        "            total += labels.size(0)\n",
        "            correct += (predicted == labels).sum().item()\n",
        "            all_labels.extend(labels.cpu().numpy())\n",
        "            all_predicted.extend(predicted.cpu().numpy())\n",
        "\n",
        "            probabilities = torch.sigmoid(outputs).cpu().numpy().flatten()  # Compute probabilities\n",
        "            all_probs.extend(probabilities)\n",
        "\n",
        "        accuracy = 100 * correct / total  # Calculate accuracy\n",
        "        cm = confusion_matrix(all_labels, all_predicted)  # Compute confusion matrix\n",
        "        tn, fp, fn, tp = cm.ravel()\n",
        "        sensitivity = 100 * tp / (tp + fn) if (tp + fn) > 0 else 0.0  # Calculate sensitivity\n",
        "        specificity = 100 * tn / (tn + fp) if (tn + fp) > 0 else 0.0  # Calculate specificity\n",
        "        auc_roc = 100 * roc_auc_score(all_labels, all_probs)  # Calculate AUC-ROC\n",
        "\n",
        "        # Print and store results\n",
        "        print(\n",
        "            f'Epoch [{epoch + 1}/{num_epochs}] Test Loss: {test_loss / len(test_dataloader):.2f} '\n",
        "            f'Test Accuracy: {accuracy:.2f}% Sensitivity: {sensitivity:.2f}% Specificity: {specificity:.2f}% AUC-ROC: {auc_roc:.4f}%'\n",
        "        )\n",
        "        results = results.append({\n",
        "            'Epoch': epoch + 1,\n",
        "            'Accuracy': accuracy,\n",
        "            'Sensitivity': sensitivity,\n",
        "            'Specificity': specificity,\n",
        "            'Test Loss': test_loss / len(test_dataloader),\n",
        "            'AUC-ROC': auc_roc\n",
        "        }, ignore_index=True)"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "---"
      ],
      "metadata": {
        "id": "WS3KlscZ1ukn"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Visualization"
      ],
      "metadata": {
        "id": "oZ5I3KUf13Gm"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "#@title #### Plotting the Accuracy, Sensitivity, Specificity, Test Loss, and AUC-ROC\n",
        "\n",
        "#plot the test metrics over each epoch\n",
        "\n",
        "# accuracy\n",
        "plt.figure(figsize=(10, 5))\n",
        "sns.regplot(x='Epoch', y='Accuracy', data=results, marker='o')\n",
        "plt.title('Accuracy Over Epochs')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.show()\n",
        "\n",
        "# sensitivity\n",
        "plt.figure(figsize=(10, 5))\n",
        "sns.regplot(x='Epoch', y='Sensitivity', data=results, marker='o', label='Sensitivity')\n",
        "plt.title('Sensitivity Over Epochs')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Sensitivity')\n",
        "plt.show()\n",
        "\n",
        "# specificity\n",
        "plt.figure(figsize=(10, 5))\n",
        "sns.regplot(x='Epoch', y='Specificity', data=results, marker='o')\n",
        "plt.title('Specificity Over Epochs')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Specificity')\n",
        "plt.show()\n",
        "\n",
        "# test-loss\n",
        "plt.figure(figsize=(10, 5))\n",
        "sns.regplot(x='Epoch', y='Test Loss', data=results, marker='o')\n",
        "plt.title('Test Loss Over Epochs')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Test Loss')\n",
        "plt.show()\n",
        "\n",
        "# auc-roc\n",
        "plt.figure(figsize=(10, 5))\n",
        "sns.regplot(x='Epoch', y='AUC-ROC', data=results, marker='o')\n",
        "plt.title('AUC-ROC Over Epochs')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('AUC-ROC')\n",
        "plt.show()\n",
        "\n",
        "\n"
      ],
      "metadata": {
        "id": "bvm6K8Ip70-0"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "fig = make_subplots(rows=2, cols=3, subplot_titles=(\"Sensitivity\", \"Accuracy\", \"AUC-ROC\", \"Specificity\", \"Test Loss\"))\n",
        "\n",
        "# Sensitivity plot\n",
        "fig.add_trace(\n",
        "    go.Scatter(x=results['Epoch'], y=results['Sensitivity'], mode='markers', marker=dict(symbol='circle', size=8)),\n",
        "    row=1, col=1\n",
        ")\n",
        "# fig.update_xaxes(title_text=\"Epoch\", row=1, col=1)\n",
        "# fig.update_yaxes(title_text=\"Sensitivity\", row=1, col=1)\n",
        "\n",
        "# Accuracy plot\n",
        "fig.add_trace(\n",
        "    go.Scatter(x=results['Epoch'], y=results['Accuracy'], mode='markers', marker=dict(symbol='circle', size=8)),\n",
        "    row=1, col=2\n",
        ")\n",
        "# fig.update_xaxes(title_text=\"Epoch\", row=1, col=2)\n",
        "# fig.update_yaxes(title_text=\"Accuracy\", row=1, col=2)\n",
        "\n",
        "# AUC ROC plot\n",
        "fig.add_trace(\n",
        "    go.Scatter(x=results['Epoch'], y=results['AUC-ROC'], mode='markers', marker=dict(symbol='circle', size=8)),\n",
        "    row=1, col=3\n",
        ")\n",
        "# fig.update_xaxes(title_text=\"Epoch\", row=1, col=3)\n",
        "# fig.update_yaxes(title_text=\"AUC ROC\", row=1, col=3)\n",
        "\n",
        "# Specificity plot\n",
        "fig.add_trace(\n",
        "    go.Scatter(x=results['Epoch'], y=results['Specificity'], mode='markers', marker=dict(symbol='circle', size=8)),\n",
        "    row=2, col=1\n",
        ")\n",
        "# fig.update_xaxes(title_text=\"Epoch\", row=2, col=1)\n",
        "# fig.update_yaxes(title_text=\"Specificity\", row=2, col=1)\n",
        "\n",
        "# Test Loss plot\n",
        "fig.add_trace(\n",
        "    go.Scatter(x=results['Epoch'], y=results['Test Loss'], mode='markers', marker=dict(symbol='circle', size=8)),\n",
        "    row=2, col=2\n",
        ")\n",
        "\n",
        "# Define different y-axis labels\n",
        "y_labels = ['True Positive Rate (TPR)', 'Accuracy', 'AUC-ROC', 'True False Rate (TFR)', 'Test Loss']\n",
        "\n",
        "for i in range(1, 6):\n",
        "    row_num = (i - 1) // 3 + 1\n",
        "    col_num = (i - 1) % 3 + 1\n",
        "\n",
        "    fig.update_xaxes(title_text=\"Epochs\", row=row_num, col=col_num)\n",
        "    fig.update_yaxes(title_text=y_labels[i-1], row=row_num, col=col_num)\n",
        "\n",
        "# Update layout\n",
        "fig.update_layout(\n",
        "    height=600,\n",
        "    width=1000,\n",
        "    title_text=\"Metrics Over Epochs\",\n",
        "    showlegend=False,\n",
        "    margin=dict(l=50, r=50, t=80, b=50)\n",
        ")\n",
        "\n",
        "# Show the plot\n",
        "fig.show()"
      ],
      "metadata": {
        "id": "aIhQDsvH9lOA"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "---"
      ],
      "metadata": {
        "id": "dRa7ndDj19tx"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Final considerations"
      ],
      "metadata": {
        "id": "aoyAW6eA2AZQ"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "As noted in the original snnTorch code, outputting the accuracy rate in this SNN model for detecting exoplanets using light curve data may not be as useful.\n",
        "\n",
        "We saw that the dataset is highly imbalanced, with a majority of examples belonging to non-exoplanets and only a few examples belonging to  exoplanets. In such cases, even a naive classifier that always predicts the majority class can achieve high accuracy. Therefore, accuracy alone may not provide a meaningful evaluation of the model's performance.\n",
        "\n",
        "Another thing to keep in mind is that this model does not perform temporal processing, which can be interesting for time-series data.\n",
        "\n",
        "Coming up, I intend to recreate this model to explore the temporal processing of a SNN model instead of processing the data in batches."
      ],
      "metadata": {
        "id": "rcnM4r_X2H1H"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "\n",
        "\n",
        "---\n",
        "\n"
      ],
      "metadata": {
        "id": "rakqi28GwErM"
      }
    },
    {
      "cell_type": "markdown",
      "source": [],
      "metadata": {
        "id": "8SHwAbQlw2Q0"
      }
    }
  ]
}