005_lstm_rnn.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model Designation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The model developed in this use case belongs to the class of non-probabilistic supervised deep learning models for solving regression tasks. Here are further explanations regarding the class designation of the model.\n",
    "\n",
    "- Supervised means that there is a reference against which performance of the model can be objectivley evaluated, within a formal protocol for evaluating models for solving regression tasks.\n",
    "\n",
    "- Deep learning means that the model will be based on the technology of artificial neural networks (ANNs), which significantly differs in comparison to the classical ML models, when it comes to the methodology of model training and testing, usage of computational resources, and most importantly, logic for making inferences which model obtains after completion of training process.\n",
    "\n",
    "- Regression means that the objective of a model is to predict numerical values (members of the set of real numbers) which originte from a continous distribution. (As opposed to classification tasks in which the objective is to predict the integers originating from discrete distribution.)\n",
    "\n",
    "- The model that will be developed is not able to (a) generate data, or (b) give an uncertianty estimate, and therefore it is designated as non-probabilistic."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Objective of Learning Task\n",
    "\n",
    "Objective of the learning task is to predict the \"temperature in kitchen area\" in degrees of Celsius, on the basis of other variables present in the \"Appliances energy prediction Data Set\" available from the UCI Machine Learning Repository. Reader is encouraged to get more information about the data set in question, having in mind that the emphasis of this use case is on developent of recurrent neural network (RNN), rather than on data analysis and data exploration."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Technology Stack\n",
    "\n",
    "Python programing language has been selected for the purposes of model development. Virtualenv package is utilized for generation of specific environment. Model development is done in Keras, while computatuion is done in TensorFlow, with Eigen as the BLAS computational backend. Data manitpulation is mostly done in numpy because there is there is no pronunced need to use named feature and target vectors. Usage of these technologies is encouraged in order to get repodocuible results."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Methodology of Model Training and Testing\n",
    "\n",
    "Modeling process follows basic methodology for training and testing of models which solve regression tasks.\n",
    "\n",
    "- The data set is split to training, validation, and testing set.\n",
    "- A simplistic LSTM RNN is constructed and trained.\n",
    "- Training process continues until no reduction of loss on validation set is recorded for two non-consecutive epoch. Minimal increase in reduction of loss, that needs to be achieved between two non-consecutive epochs, is 0.01 MSE.\n",
    "- After completion of the training process, model performance is evaluated on the test set. Main component of model performance is its accuracy, which is measured as mean absolute error (MAE). Model generalization ability, as the second component of the model performance, is only briefly evaluated.\n",
    "- Finally, alongside the evaluation of model accuracy, a comparison is done between a fit produces by a FFD ANN developed previousley, and LSTM RNN.\n",
    "\n",
    "No cross-validation or model fine tuning is performed. The reasoning behind such a simplistic approach to model training and testing, is driven by limitations of the model development environment, especially in the terms of computational resources. Therefore, methodology of model training and testing is designed to meet the requirements of fast development and low computational effort, rather than designed to find best solution in terms of model's generalization ability and accuracy. Model training and testing proces would greatly benefit from introduction of hyper-parameter tuning and cross validation.\n",
    "\n",
    "Also, it is important to take into account that the main purpose of the use-case is to gain insight into preformance of LSTM RNN in comparison to the FFD ANN. Having in mind that the time series data is used in the use-case, expected outcome is to observe advantages of LSTM RNN over FFD ANNs.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model Development"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "First step is to import necessary packages."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from math import floor\n",
    "from collections import OrderedDict\n",
    "from keras.models import load_model\n",
    "from keras.layers import Input, Dense, LSTM\n",
    "from keras.optimizers import Adam\n",
    "from keras.models import Model\n",
    "from keras.callbacks import EarlyStopping\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.gridspec as grid"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now, we proceed with declaration of a class for data holding the names of the features and targets. This class also preforms basic data preparation, hence the name of the class. However, further preparation of the data is performed downstream as well."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [],
   "source": [
    "class PreparedData:\n",
    "    def __init__(self, inn):\n",
    "        self.original_data = inn\n",
    "        self.prepared_data = None\n",
    "        self.feature_labels = None\n",
    "        self.target_labels = None\n",
    "        # Assign index and convert to date.\n",
    "        out = inn.copy(deep=True)\n",
    "        out.index = pd.to_datetime(inn['date'])\n",
    "        # Drop unnecessary columns.\n",
    "        out.drop(columns=['date', 'rv1', 'rv2'], inplace=True)\n",
    "        # Get labels.\n",
    "        target_labels = [\n",
    "            label for label in inn.columns if\n",
    "            ('T1' in label)\n",
    "            & ('_out' not in label)\n",
    "            & ('Tdewpoint' not in label)\n",
    "        ]\n",
    "        feature_labels = set(out.columns) - set(target_labels)\n",
    "        self.prepared_data = out\n",
    "        self.feature_labels = feature_labels\n",
    "        self.target_labels = target_labels"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The purpose of the following function is to create a subset of data by selecting desired columns."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "outputs": [],
   "source": [
    "def subset_columns(frame, columns):\n",
    "    \"\"\"\n",
    "    Return numpy array wh. subset of columns.\n",
    "    \"\"\"\n",
    "    return np.array(frame.loc[:, columns])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Previous function is then followed with two utility functions. One, to make feature data sets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {},
   "outputs": [],
   "source": [
    "def make_features(training_frame,\n",
    "                  validation_frame,\n",
    "                  testing_frame,\n",
    "                  feature_columns):\n",
    "    \"\"\"\n",
    "    Return a features for training, validation, and testing.\n",
    "    \"\"\"\n",
    "    return (\n",
    "        subset_columns(\n",
    "            frame=training_frame,\n",
    "            columns=feature_columns,\n",
    "        ),\n",
    "        subset_columns(\n",
    "            frame=validation_frame,\n",
    "            columns=feature_columns,\n",
    "        ),\n",
    "        subset_columns(\n",
    "            frame=testing_frame,\n",
    "            columns=feature_columns\n",
    "        )\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And another to make targets data sets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {},
   "outputs": [],
   "source": [
    "def make_targets(training_frame,\n",
    "                 validation_frame,\n",
    "                 testing_frame,\n",
    "                 targets_columns):\n",
    "    \"\"\"\n",
    "    Return a features for training, validation, and testing.\n",
    "    \"\"\"\n",
    "    return (\n",
    "        subset_columns(\n",
    "            frame=training_frame,\n",
    "            columns=targets_columns,\n",
    "        ),\n",
    "        subset_columns(\n",
    "            frame=validation_frame,\n",
    "            columns=targets_columns,\n",
    "        ),\n",
    "        subset_columns(\n",
    "            frame=testing_frame,\n",
    "            columns=targets_columns\n",
    "        )\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Utility funcitions are complemented with the function to standardize data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {},
   "outputs": [],
   "source": [
    "def compute_mean_and_standard_deviation(table):\n",
    "    \"\"\"\n",
    "    Return mean and standard deviation of the table.\n",
    "    \"\"\"\n",
    "    return np.mean(table, axis=0), np.std(table, axis=0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we get into functions specific for the usage of RNNs. First in line, is a function that reshapes an array into a shape that can be ingested into a RNN (samples, time-steps, dimensions). In the vocabulary of the function, the length of the sequence (number of the time-steps) is designated as \"memory\", in order to emphaisise the capacity of the network to absorb temporal data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [],
   "source": [
    "def reshape_features_for_rnn(input_array, memory):\n",
    "    \"\"\"\n",
    "    Return features reshaped to the required dimensionality.\n",
    "    \"\"\"\n",
    "    feasible_duration_of_training_set = len(input_array[memory:, :])\n",
    "    # Number of samples.\n",
    "    total_samples = floor(feasible_duration_of_training_set / memory)\n",
    "    # Subset features, so that there is enough timestamps for all samples.\n",
    "    # and return with the new shape.\n",
    "    return np.reshape(\n",
    "        input_array[0:memory*total_samples, :],\n",
    "        newshape=(\n",
    "            total_samples, memory, np.shape(input_array)[1]\n",
    "        )\n",
    "    )\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we go into a series of generators which yield baches. The first in line is the generator for the training process. This generator yields batches of featurea and targets. For the generaor to work it is necessary to firstly create an inventory of batches, which will be passed to the generator as an input. Production of individual batches is done via call to the function defined above."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 144,
   "metadata": {},
   "outputs": [],
   "source": [
    "def generate_batch(features, labels, inventory_of_batches, samples_per_batch):\n",
    "    \"\"\"\n",
    "    Return a batch of features and targets of desired size.\n",
    "    \"\"\"\n",
    "    while True:\n",
    "        for batch in inventory_of_batches.index:\n",
    "            feature_sample_start = \\\n",
    "                inventory_of_batches['features_start'][batch]\n",
    "            feature_sample_end = \\\n",
    "                inventory_of_batches['features_end'][batch]\n",
    "            target_sample_start = \\\n",
    "                inventory_of_batches['targets_start'][batch]\n",
    "            target_sample_end = \\\n",
    "                inventory_of_batches['targets_end'][batch]\n",
    "            features_batch = np.reshape(\n",
    "                features[feature_sample_start:feature_sample_end, :],\n",
    "                newshape=(\n",
    "                    samples_per_batch,\n",
    "                    memory,\n",
    "                    np.shape(x_test_standardized)[1]\n",
    "                )\n",
    "            )\n",
    "            targets_batch = np.reshape(\n",
    "                 labels[target_sample_start:target_sample_end, ],\n",
    "                 newshape=(samples_per_batch, 1)\n",
    "            )\n",
    "            yield (\n",
    "                {'input_layer': features_batch},\n",
    "                {'output_layer': targets_batch}\n",
    "            )\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Next in line is the generator of the batches for the prediction process. What is the difference between this generator, and the first one? This generator yields only features, and no targets. Everything else is the same."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [],
   "source": [
    "def generate_batch_for_prediction(features, inventory_of_batches):\n",
    "    \"\"\"\n",
    "    Return a batch of features and targets of desired size.\n",
    "    \"\"\"\n",
    "    while True:\n",
    "        for batch in inventory_of_batches.index:\n",
    "            feature_sample_start = \\\n",
    "                inventory_of_batches['features_start'][batch]\n",
    "            feature_sample_end = \\\n",
    "                inventory_of_batches['features_end'][batch]\n",
    "            features_batch = np.reshape(\n",
    "                features[feature_sample_start:feature_sample_end, :],\n",
    "                newshape=(1, memory, np.shape(x_test_standardized)[1])\n",
    "            )\n",
    "            yield ({'input_layer': features_batch})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now there are only two things left. First one is to create a function to make the LSMT RNN."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_lstm_rnn_model(capacity, dimensionality):\n",
    "    \"\"\"\n",
    "    Return model.\n",
    "    \"\"\"\n",
    "    # Declare input layer.\n",
    "    input_layer = Input(shape=(capacity, dimensionality), name='input_layer')\n",
    "    # Declare regression layer 1.\n",
    "    regression_1 = LSTM(\n",
    "        units=2,\n",
    "        unit_forget_bias=True,\n",
    "        return_sequences=True,\n",
    "        stateful=False,\n",
    "        implementation=1\n",
    "    )(input_layer)\n",
    "    # Declare regression layer 2.\n",
    "    regression_2 = LSTM(\n",
    "        units=2,\n",
    "        unit_forget_bias=True,\n",
    "        return_sequences=False,\n",
    "        stateful=False,\n",
    "        implementation=1\n",
    "    )(regression_1)\n",
    "    output = Dense(1, activation='linear', name='output_layer')(regression_2)\n",
    "    return Model(inputs=input_layer, outputs=output)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And also to create a function the generate predictions form the FFD ANN."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_stochastic_predictions(model_for_deployment, input_data, passes):\n",
    "    \"\"\"\n",
    "    Return stochastic predictions according to desired number of passes.\n",
    "    \"\"\"\n",
    "    # Declare a list to store predictions.\n",
    "    raw_predictions = []\n",
    "    for forward_pass in range(passes):\n",
    "        raw_predictions.append(\n",
    "            model_for_deployment.predict(\n",
    "                x=input_data, batch_size=len(input_data), verbose=0)\n",
    "        )\n",
    "    return np.transpose(np.reshape(\n",
    "        raw_predictions, newshape=(passes, len(input_data))\n",
    "    ))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Without further ado, let us move on to the mode development."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [],
   "source": [
    "if __name__ == '__main__':\n",
    "\n",
    "    # Declare file handles.\n",
    "    RAW_DATA = r'M:\\Projects\\004_mc_dropout\\1_data\\energydata_complete.csv'\n",
    "    FFD_ANN_FOLDER = r'M:\\Projects\\004_mc_dropout\\3_output'\n",
    "\n",
    "    # Get data.\n",
    "    raw_data = pd.read_csv(RAW_DATA, sep=',')\n",
    "\n",
    "    # Prepare data.\n",
    "    data_preparation = PreparedData(inn=raw_data)\n",
    "    prepared_data = data_preparation.prepared_data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Firstly, the data is imported and prepared. Then, we declare start and end timestamps for training, validation, and testing sets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Declare limits for subsets of data set.\n",
    "    training_start = 0\n",
    "    training_end = int(len(prepared_data)/2 - len(prepared_data)/2*.1)\n",
    "    validation_start = int(len(prepared_data)/2 - len(prepared_data)/2*.1)\n",
    "    validation_end = int(len(prepared_data)/2)\n",
    "    testing_start = int(len(prepared_data)/2)\n",
    "    testing_end = len(prepared_data)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Training data is then subsetted.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Subset samples for training.\n",
    "    features_and_targets_for_training = \\\n",
    "        prepared_data.iloc[training_start:training_end, :]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And here we can see the start, and end timestamps."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Start of training: 2016-01-11 17:00:00. End of training: 2016-03-13 08:50:00.\n"
     ]
    }
   ],
   "source": [
    "    # Print starting and ending times.\n",
    "    print('Start of training: %s. End of training: %s.' % (\n",
    "        features_and_targets_for_training.index[0],\n",
    "        features_and_targets_for_training.index[-1]\n",
    "    ))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we can also subset the validation and testing data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Subset samples for validation.\n",
    "    features_and_targets_for_validation = \\\n",
    "        prepared_data.iloc[validation_start:validation_end, :]\n",
    "\n",
    "    # Subset samples for testing.\n",
    "    features_and_targets_for_testing = \\\n",
    "        prepared_data.iloc[testing_start:testing_end, :]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And examine start time-stamps and end time-stapms of these sections of the data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Start of validation: 2016-03-13 09:00:00. End of validation: 2016-03-20 05:20:00.\n",
      "Start of testing: 2016-03-20 05:30:00. End of testing: 2016-05-27 18:00:00.\n"
     ]
    }
   ],
   "source": [
    "    print('Start of validation: %s. End of validation: %s.' % (\n",
    "        features_and_targets_for_validation.index[0],\n",
    "        features_and_targets_for_validation.index[-1]\n",
    "    ))\n",
    "    print('Start of testing: %s. End of testing: %s.' % (\n",
    "        features_and_targets_for_testing.index[0],\n",
    "        features_and_targets_for_testing.index[-1]\n",
    "    ))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As mentioned in the introduction, the purpose of the use case is to make comparison betwenn FFD ANN and LSTM RNN. Therefore, we firts need to generate predictions from the FFD ANN model, and to do that we need to prepare the data in the same way that the data was prepared for the training of FFD ANN. So we will start with loading of the scaling parameters."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Get FFD ANN model scaling parameters.\n",
    "    ffd_ann_scaling_parameters = pd.read_csv(\n",
    "        filepath_or_buffer=\n",
    "            FFD_ANN_FOLDER+r'\\uc004_mc_dropout_scaling_parameters.csv',\n",
    "        index_col=0\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Then, we will make features and targets but with special consideration to the order of the variables. That is the purpose of the reference to \"variable_names\" column of the scaling parmeters data frame. This column, holds the order of the variables."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Get features.\n",
    "    features_training, features_validation, features_testing = \\\n",
    "        make_features(\n",
    "            training_frame=features_and_targets_for_training,\n",
    "            validation_frame=features_and_targets_for_validation,\n",
    "            testing_frame=features_and_targets_for_testing,\n",
    "            feature_columns=ffd_ann_scaling_parameters['variable_name'],\n",
    "        )\n",
    "\n",
    "    # Get targets.\n",
    "    targets_train, targets_validation, targets_test = make_targets(\n",
    "        training_frame=features_and_targets_for_training,\n",
    "        validation_frame=features_and_targets_for_validation,\n",
    "        testing_frame=features_and_targets_for_testing,\n",
    "        targets_columns=data_preparation.target_labels,\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now all that there is left is to get the actual scaling parameters (mean, and standard deviation) and to standardize the data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Get mean and standard deviation of training features.\n",
    "    mean = ffd_ann_scaling_parameters['mean'].values\n",
    "    standard_deviation = \\\n",
    "        ffd_ann_scaling_parameters['standard_deviation'].values\n",
    "    # Standardize training, validation, and testing features.\n",
    "    x_train_standardized = (features_training-mean) / standard_deviation\n",
    "    x_validation_standardized = (features_validation-mean) / standard_deviation\n",
    "    x_test_standardized = (features_testing-mean) / standard_deviation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let us verify if the standardization was done correclty."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1.]\n",
      "[1.18307181 0.6743218  0.16078519 1.3665739  0.20565271 0.54285796\n",
      " 1.15865454 0.87941373 0.27765162 0.06666029 0.0703394  0.41331868\n",
      " 0.65862733 0.10223157 0.3124564  0.74120142 1.93688439 0.25435944\n",
      " 0.75422202 0.04869117 1.09058073 0.30501772 0.58681752 0.41685807\n",
      " 0.16557348]\n",
      "[0.78632384 2.10412644 1.85421244 2.09215122 0.64130847 1.73711587\n",
      " 2.28147666 0.41874729 0.71785236 0.369263   0.62236422 0.52631285\n",
      " 0.76336401 0.81455222 1.12648601 2.5185316  0.536103   0.92568494\n",
      " 1.70935316 2.49731804 1.6243543  1.36839702 1.75868663 0.60882017\n",
      " 0.61950162]\n"
     ]
    }
   ],
   "source": [
    "    # Print variance of the features.\n",
    "    print(np.var(x_train_standardized, axis=0))\n",
    "    print(np.var(x_validation_standardized, axis=0))\n",
    "    print(np.var(x_test_standardized, axis=0))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As it can be seen, the standardization is done correctly. The shift in distribution of the input data is already known characteristic of this data set. Now, we get the FFD ANN."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Get FFD ANN model.\n",
    "    ffd_ann_model = load_model(\n",
    "        filepath=FFD_ANN_FOLDER+r'\\uc004_mc_dropout_model.h5'\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And generate some predictions out of FFD ANN."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {},
   "outputs": [],
   "source": [
    "    ffd_ann_predictions = get_stochastic_predictions(\n",
    "        model_for_deployment=ffd_ann_model,\n",
    "        input_data=x_test_standardized,\n",
    "        passes=30\n",
    "    )\n",
    "    ffd_ann_predictions = np.mean(ffd_ann_predictions, axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let us recall what was the accracy of the FFD ANN."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.75\n"
     ]
    }
   ],
   "source": [
    "     # Compute MAE.\n",
    "    ffd_ann_mae = np.round(\n",
    "        np.mean(np.abs(targets_test.flatten()-ffd_ann_predictions)),\n",
    "        decimals=2\n",
    "    )\n",
    "    print(ffd_ann_mae)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Not a bad result 0.75 Celsius. Let us now move on to set-up LSTM RNN. Firstly, we need to declare how much will be the memory (ie. storage) capacity of the LSTM RNN. We will set it to 30 time-steps."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Declare how much memory is needed for prediction?\n",
    "    memory = 30"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The next thing to declare is the prediction interval, or how often we need to make predictions. Let assume that predictions need to be made at every timestep."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Declare prediction interval.\n",
    "    prediction_interval = 1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Another name for the prediction interval, cold be prediction frequency. Apart, from predicion interval, we also need the length of prediction, or how many time-steps will be predicted every time the model is executed."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Prediction length.\n",
    "    prediction_length = 1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We declare prediction length to 1 time-step. Now, how we can state the problem the network is trying to solve? Given the last 30 time-steps of data in the features (memory), predict the \"temperature int the kitchen area\" at the next (31st) time-step (prediction length), once per every time-step (prediction interval). Please note that there is no delay between the memory and prediction. As soon as the 30 time-steps is consumed, prediction for the 31st time-step is made. (The problem could be specified with a longer delay.) Also note that there is no down-sampling of the input features. (All 30 time-steps are taken into account when performing prediction). Finally, the length of predicion could be increased, to more than one. Now, we need to declare some paramaters which will be utilized in training."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Batch size.\n",
    "    samples_per_batch = 1\n",
    "    batch_length = samples_per_batch * memory"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We need to declare how many samples will be witin the batch. In this case each batch will contain only one sample. Batch length is just a total ength of each batch given the number of samples within the batch and memory capacity of the network. Again, this is one of the tuning parameters, or better yet--design decisions--on how to construct model training and deployment process. Please, also note that in this setup introduction of more than one sample within a batch will mean that there will be overalapping of the time-steps among the batches. After decalring these basic parameters of model deployment and training process, we need to determine derived inputs for the trining process."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Compute maximal number of samples.\n",
    "    training_samples = len(x_train_standardized) - batch_length\n",
    "    validation_samples = len(x_validation_standardized) - batch_length\n",
    "    testing_samples = len(x_test_standardized) - batch_length"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finally, we build an inventory of the training indices which will utilized by the fit generator to yield batches of samples for the features and targets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Indices for training.\n",
    "    training_indices_features = pd.DataFrame(OrderedDict([\n",
    "        (\n",
    "            'features_start',\n",
    "            list(range(0, training_samples, samples_per_batch))\n",
    "        ), (\n",
    "            'features_end',\n",
    "            list(range(\n",
    "                batch_length,\n",
    "                batch_length+training_samples,\n",
    "                samples_per_batch\n",
    "            ))\n",
    "        ), (\n",
    "            'targets_start',\n",
    "            list(range(\n",
    "                batch_length,\n",
    "                batch_length+training_samples,\n",
    "                samples_per_batch\n",
    "            ))\n",
    "        ), (\n",
    "            'targets_end',\n",
    "            list(range(\n",
    "                batch_length+samples_per_batch,\n",
    "                batch_length+samples_per_batch+training_samples,\n",
    "                samples_per_batch\n",
    "            ))\n",
    "        )\n",
    "    ]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Befor we complete this inventory, we need to make sure that there are no implausible indices."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Check.\n",
    "    training_indices_features['check'] = \\\n",
    "        np.where(\n",
    "            training_indices_features['targets_end'] > len(targets_train), 1, 0\n",
    "        )\n",
    "    training_indices_features = training_indices_features.loc[\n",
    "        training_indices_features['check'] == 0, :\n",
    "    ]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "At the very end we will shuffle the indices. Please note that (a) only the indices are shuffled, and not the arrays that contain data, and that (b) the whole batches are shuffled and not the individual samples within the batches. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Shuffle training indices.\n",
    "    shuffle = np.random.randint(\n",
    "        low=0,\n",
    "        high=len(training_indices_features),\n",
    "        size=len(training_indices_features),\n",
    "        dtype=np.int\n",
    "    )\n",
    "    training_indices_features = training_indices_features.iloc[shuffle, :]\n",
    "    training_indices_features.reset_index(inplace=True, drop=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "With the exception of shuffling we repeat this indexing process for both validation data, and testing data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Validation indices.\n",
    "    validation_indices = pd.DataFrame(OrderedDict([\n",
    "        (\n",
    "            'features_start',\n",
    "            list(range(0, validation_samples, prediction_length))\n",
    "        ), (\n",
    "            'features_end',\n",
    "            list(range(\n",
    "                batch_length,\n",
    "                batch_length+validation_samples,\n",
    "                prediction_length\n",
    "            ))\n",
    "        ), (\n",
    "            'targets_start',\n",
    "            list(range(\n",
    "                batch_length,\n",
    "                batch_length+validation_samples,\n",
    "                prediction_length\n",
    "            )),\n",
    "        ), (\n",
    "            'targets_end',\n",
    "            list(range(\n",
    "                batch_length+samples_per_batch,\n",
    "                batch_length+samples_per_batch+validation_samples,\n",
    "                prediction_length\n",
    "            ))\n",
    "        )\n",
    "    ]))\n",
    "\n",
    "    # Check.\n",
    "    validation_indices['check'] = \\\n",
    "        np.where(\n",
    "            validation_indices['targets_end'] > len(targets_validation), 1, 0\n",
    "        )\n",
    "    validation_indices = validation_indices.loc[\n",
    "        validation_indices['check'] == 0, :\n",
    "    ]\n",
    "\n",
    "    # Testing indices.\n",
    "    testing_indices = pd.DataFrame(OrderedDict([\n",
    "        (\n",
    "            'features_start',\n",
    "            list(range(0, testing_samples, prediction_length))\n",
    "        ), (\n",
    "            'features_end',\n",
    "            list(range(\n",
    "                memory,\n",
    "                memory+testing_samples,\n",
    "                prediction_length\n",
    "            ))\n",
    "        ), (\n",
    "            'targets_start',\n",
    "            list(range(\n",
    "                memory,\n",
    "                memory+testing_samples,\n",
    "                prediction_length\n",
    "            ))\n",
    "        ), (\n",
    "            'targets_end',\n",
    "            list(range(\n",
    "                memory+samples_per_batch,\n",
    "                memory+samples_per_batch+testing_samples,\n",
    "                prediction_length\n",
    "            ))\n",
    "        )\n",
    "    ]))\n",
    "\n",
    "    # Check.\n",
    "    testing_indices['check'] = \\\n",
    "        np.where(\n",
    "            testing_indices['features_end'] > len(x_test_standardized), 1, 0\n",
    "        )\n",
    "    testing_indices = testing_indices.loc[testing_indices['check'] == 0, :]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In order to provide information to the batch generator when to stop, we need to compute the number of steps for training, validation, and testing."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Steps per epoch.\n",
    "    training_steps = len(training_indices_features)\n",
    "    validation_steps = len(validation_indices)\n",
    "    testing_steps = len(testing_indices)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And this puts an end to the preparations. Now, we can make a model with a call to the function."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_layer (InputLayer)     (None, 30, 25)            0         \n",
      "_________________________________________________________________\n",
      "lstm_3 (LSTM)                (None, 30, 2)             224       \n",
      "_________________________________________________________________\n",
      "lstm_4 (LSTM)                (None, 2)                 40        \n",
      "_________________________________________________________________\n",
      "output_layer (Dense)         (None, 1)                 3         \n",
      "=================================================================\n",
      "Total params: 267\n",
      "Trainable params: 267\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "    # Get LSTM RNN.\n",
    "    lstm_rnn_model = get_lstm_rnn_model(\n",
    "        capacity=memory, dimensionality=np.shape(x_train_standardized)[1]\n",
    "    )\n",
    "\n",
    "    # Print model summary.\n",
    "    print(lstm_rnn_model.summary())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As it can be seen the model was declared correctly. Then networks consists of two inner layers with two LSTM cells in each layer, and an output layer with one unit, and linear activation. Let us declare an optimizer."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Declare optimizer.\n",
    "    optimizer = Adam(\n",
    "        lr=0.001,\n",
    "        beta_1=0.9,\n",
    "        beta_2=0.999,\n",
    "        epsilon=None,\n",
    "        decay=0.0,\n",
    "        amsgrad=False\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Nothing special here. Adam optimizer with default parameters. All that is left is to compile the model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Compile model.\n",
    "    lstm_rnn_model.compile(\n",
    "        optimizer=optimizer,\n",
    "        loss='mse',\n",
    "        metrics=['mae']\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And of we go with the training process. Training process will go for up to 50 epoch, and will discontinue according to criteria defined in the callback. After the training process is discontinued, the callback will restore the best model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "8850/8850 [==============================] - 364s 41ms/step - loss: 85.1369 - mean_absolute_error: 7.3497 - val_loss: 1.8311 - val_mean_absolute_error: 1.1635\n",
      "Epoch 2/50\n",
      "8850/8850 [==============================] - 349s 39ms/step - loss: 1.0268 - mean_absolute_error: 0.7539 - val_loss: 0.3659 - val_mean_absolute_error: 0.4920\n",
      "Epoch 3/50\n",
      "8850/8850 [==============================] - 338s 38ms/step - loss: 0.2479 - mean_absolute_error: 0.3584 - val_loss: 0.2656 - val_mean_absolute_error: 0.3840\n",
      "Epoch 4/50\n",
      "8850/8850 [==============================] - 383s 43ms/step - loss: 0.0720 - mean_absolute_error: 0.2016 - val_loss: 0.2734 - val_mean_absolute_error: 0.3809\n",
      "Epoch 5/50\n",
      "8850/8850 [==============================] - 323s 37ms/step - loss: 0.0498 - mean_absolute_error: 0.1712 - val_loss: 0.2502 - val_mean_absolute_error: 0.3537\n",
      "Epoch 6/50\n",
      "8850/8850 [==============================] - 364s 41ms/step - loss: 0.0445 - mean_absolute_error: 0.1619 - val_loss: 0.2408 - val_mean_absolute_error: 0.3434\n",
      "Epoch 7/50\n",
      "8850/8850 [==============================] - 422s 48ms/step - loss: 0.0413 - mean_absolute_error: 0.1566 - val_loss: 0.2457 - val_mean_absolute_error: 0.3477\n",
      "Restoring model weights from the end of the best epoch\n",
      "Epoch 00007: early stopping\n"
     ]
    }
   ],
   "source": [
    "    # Fit model.\n",
    "    history = lstm_rnn_model.fit_generator(\n",
    "        generate_batch(\n",
    "            features=x_train_standardized,\n",
    "            labels=targets_train,\n",
    "            inventory_of_batches=training_indices_features,\n",
    "            samples_per_batch=samples_per_batch\n",
    "        ),\n",
    "        steps_per_epoch=training_steps,\n",
    "        epochs=50,\n",
    "        validation_data=generate_batch(\n",
    "            features=x_validation_standardized,\n",
    "            labels=targets_validation,\n",
    "            inventory_of_batches=validation_indices,\n",
    "            samples_per_batch=samples_per_batch\n",
    "        ),\n",
    "        validation_steps=validation_steps,\n",
    "        callbacks=[\n",
    "            EarlyStopping(\n",
    "                monitor='val_loss',\n",
    "                min_delta=0.01,\n",
    "                patience=2,\n",
    "                restore_best_weights=True,\n",
    "                verbose=1\n",
    "            )\n",
    "        ]\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Even though it was a short training process let us take a look at how it went."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Loss')"
      ]
     },
     "execution_count": 150,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 960.001x540 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "    # Plot accuracy and loss.\n",
    "    fig = plt.figure(figsize=[26.6667*.5, 15*.5])\n",
    "    plotting_grid = grid.GridSpec(2, 1)\n",
    "    # Plot accuracy.\n",
    "    ax1 = fig.add_subplot(plotting_grid[0, 0])\n",
    "    ax1.plot(history.history['mean_absolute_error'], label='Training')\n",
    "    ax1.plot(history.history['val_mean_absolute_error'], label='Validation')\n",
    "    ax1.set_xticks(list(range(\n",
    "        0, len(history.history['val_mean_absolute_error'])\n",
    "    )))\n",
    "    ax1.invert_yaxis()\n",
    "    ax1.legend()\n",
    "    ax1.set_title('Accuracy')\n",
    "    # Plot loss.\n",
    "    ax2 = fig.add_subplot(plotting_grid[1, 0])\n",
    "    ax2.plot(history.history['loss'], label='Training')\n",
    "    ax2.plot(history.history['val_loss'], label='Validation')\n",
    "    ax2.set_xticks(list(range(\n",
    "        0, len(history.history['val_mean_absolute_error'])\n",
    "    )))\n",
    "    ax2.legend()\n",
    "    ax2.set_title('Loss')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now, let us generate preictions out of LSTM RNN."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 151,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "9838/9838 [==============================] - 91s 9ms/step\n"
     ]
    }
   ],
   "source": [
    "    # Generate predictions.\n",
    "    lstm_rnn_predictions = lstm_rnn_model.predict_generator(\n",
    "        generator=generate_batch_for_prediction(\n",
    "            features=x_test_standardized,\n",
    "            inventory_of_batches=testing_indices\n",
    "        ),\n",
    "        steps=len(testing_indices),\n",
    "        verbose=1\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Just a short check, and a remainder tht we are dropping first thiry predictions becauase of design of RNN network."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 152,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "False\n",
      "True\n"
     ]
    }
   ],
   "source": [
    "    print(len(lstm_rnn_predictions) == len(targets_test))\n",
    "    print(len(lstm_rnn_predictions) == len(targets_test[30:, ]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now, we can put all predictions, both from FFD ANN and LSTM RNN into a single data frame."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 153,
   "metadata": {},
   "outputs": [],
   "source": [
    "    table_for_plotting = pd.DataFrame(OrderedDict([\n",
    "        ('targets', targets_test[30:].flatten()),\n",
    "        ('ffd_ann_predictions', ffd_ann_predictions[30:, ]),\n",
    "        ('lstm_rnn_predictions', lstm_rnn_predictions.flatten())\n",
    "    ]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let us now compute the accuracy of the LSTM RNN."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # Compute accuracy (MAE).\n",
    "    lstm_rnn_mae = np.round(\n",
    "        np.mean(\n",
    "            np.abs(\n",
    "                table_for_plotting['targets']\n",
    "                - table_for_plotting['lstm_rnn_predictions']\n",
    "            )\n",
    "        ),\n",
    "        decimals=2\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now let us compare the fit of these two distinct types of networks."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1ce571fcc50>"
      ]
     },
     "execution_count": 155,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 960.001x540 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "    # Comparison of fit.\n",
    "    fig = plt.figure(figsize=[26.6667 * .5, 15 * .5])\n",
    "    fig.subplots_adjust(hspace=.25)\n",
    "    plotting_grid = grid.GridSpec(2, 1)\n",
    "    # Plot LSTM RNN.\n",
    "    ax1 = fig.add_subplot(plotting_grid[0, 0])\n",
    "    ax1.plot(\n",
    "        table_for_plotting['targets'],\n",
    "        color='orangered',\n",
    "        label='Targets',\n",
    "        linestyle=':',\n",
    "        linewidth=1,\n",
    "    )\n",
    "    ax1.plot(\n",
    "        table_for_plotting['lstm_rnn_predictions'],\n",
    "        color='dodgerblue',\n",
    "        label='LSTM RNN predictions',\n",
    "        linewidth=1\n",
    "    )\n",
    "    ax1.set_title(\n",
    "        'Fit of LSTM RNN\\n(MAE=%s)'\n",
    "        %\n",
    "        str(lstm_rnn_mae)\n",
    "    )\n",
    "    ax1.legend(loc=1)\n",
    "    # Plot FFD ANN.\n",
    "    ax2 = fig.add_subplot(plotting_grid[1, 0])\n",
    "    ax2.plot(\n",
    "        table_for_plotting['targets'],\n",
    "        color='orangered',\n",
    "        label='Targets',\n",
    "        linestyle=':',\n",
    "        linewidth=1,\n",
    "    )\n",
    "    ax2.plot(\n",
    "        table_for_plotting['ffd_ann_predictions'],\n",
    "        color='dodgerblue',\n",
    "        label='FFD ANN predictions',\n",
    "        linewidth=1\n",
    "    )\n",
    "    ax2.set_title(\n",
    "        'Fit of FFD ANN\\n(MAE=%s)'\n",
    "        %\n",
    "        str(ffd_ann_mae)\n",
    "    )\n",
    "    ax2.legend(loc=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Conclusion\n",
    "\n",
    "As we can see, there is obviousley a distinctive, much more suitable fit, given by the RNN with LSTM cells, in comparison to the FFD ANN. This is a consequence of a specific architecture of the RNN, which allows the network to become more skilfull in making conclusions, which take into account temporal dependance. In contrast, FFD ANN is completley unaware of temporal dependance in the data. FFD ANN makes concusions only on the basis of the inputs which are completely synchronized with the outputs in temporeal dimension, esentially downgrading time series data into transactional (i.i.d.) data. The advantages do not stop there, relative improvement in the accuracy over FFD ANN is 21%. The RNN was able to reduce MAE for 0.16 degrees of Celsius, which is one fith of the result achieved by the FFD ANN. Also, please note that the FFD ANN is heavily regularized, while there is no regularization of the LSTM RNN. However, the analysis would not be complete without two additional notes. First one is a note regarding the training time, which is several times greater when it comes to the LSTM RNN in comparison with FFD ANN. FFD ANNs fit quickly, and require no tuning of the dimension of time. Second one is the ability of FFD ANN to give a slightly better fit in the second part of the data, after there is a shift in the distribution of the input data. In this section of the data LSTM RNN is demonstrating lower fit ability, most likely because of absence of any reqularization during training process."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}