CORTEX-scVI-stability.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# loading modules"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "%matplotlib inline\n",
    "\n",
    "from model import scVI_final as scVI\n",
    "from benchmarking import *\n",
    "from helper import *"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "learning_rate = 0.0004\n",
    "epsilon = 0.01"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# import data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# expression data\n",
    "data_path = \"/home/ubuntu/single-cell-scVI/data/Zeisel/\"\n",
    "expression_train = np.loadtxt(data_path + \"data_train\")\n",
    "expression_test = np.loadtxt(data_path + \"data_test\")\n",
    "\n",
    "# zero masked matrix\n",
    "X_zero, i, j, ix = \\\n",
    "        np.load(data_path + \"imputation/X_zero.npy\"), np.load(data_path + \"imputation/i.npy\"),\\\n",
    "        np.load(data_path + \"imputation/j.npy\"), np.load(data_path + \"imputation/ix.npy\")\n",
    "#labels\n",
    "c_train = np.loadtxt(data_path + \"label_train\")\n",
    "c_test = np.loadtxt(data_path + \"label_test\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Computational graph"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# getting priors\n",
    "log_library_size = np.log(np.sum(expression_train, axis=1))\n",
    "mean, var = np.mean(log_library_size), np.var(log_library_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running scVI on 558 genes\n",
      "scVI will run without batch correction\n",
      "Will work on mode numeric for incorporating library size\n",
      "Will work on mode gene for modeling inverse dispersion param\n",
      "Will apply zero inflation\n",
      "1 hidden layers at 128 each for a final 2 latent space\n",
      "Running scVI on 558 genes\n",
      "scVI will run without batch correction\n",
      "Will work on mode numeric for incorporating library size\n",
      "Will work on mode gene for modeling inverse dispersion param\n",
      "Will apply zero inflation\n",
      "1 hidden layers at 128 each for a final 2 latent space\n",
      "Running scVI on 558 genes\n",
      "scVI will run without batch correction\n",
      "Will work on mode numeric for incorporating library size\n",
      "Will work on mode gene for modeling inverse dispersion param\n",
      "Will apply zero inflation\n",
      "1 hidden layers at 128 each for a final 2 latent space\n",
      "Running scVI on 558 genes\n",
      "scVI will run without batch correction\n",
      "Will work on mode numeric for incorporating library size\n",
      "Will work on mode gene for modeling inverse dispersion param\n",
      "Will apply zero inflation\n",
      "1 hidden layers at 128 each for a final 2 latent space\n"
     ]
    }
   ],
   "source": [
    "res = np.zeros((5, 6, 3))\n",
    "latent_d = [2, 5, 7, 10, 15, 20]\n",
    "for j, d in enumerate(latent_d):\n",
    "    for t in range(5):\n",
    "        tf.reset_default_graph()\n",
    "        expression = tf.placeholder(tf.float32, (None, expression_train.shape[1]), name='x')\n",
    "        kl_scalar = tf.placeholder(tf.float32, (), name='kl_scalar')\n",
    "        optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate, epsilon=epsilon)\n",
    "        training_phase = tf.placeholder(tf.bool, (), name='training_phase')\n",
    "        model = scVI.scVIModel(expression=expression, kl_scale=kl_scalar, \\\n",
    "                                 optimize_algo=optimizer, phase=training_phase, \\\n",
    "                                  library_size_mean=mean, library_size_var=var, n_latent=d)\n",
    "\n",
    "        # Session creation\n",
    "        sess = tf.Session()\n",
    "        sess.run(tf.global_variables_initializer())\n",
    "\n",
    "        result = train_model(model, (expression_train, expression_test), sess, 250)\n",
    "\n",
    "        dic_full = {expression: expression_train, training_phase:False}\n",
    "        latent = sess.run(model.z, feed_dict=dic_full)\n",
    "        res[t, j] = cluster_scores(latent, 7, c_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEKCAYAAAD9xUlFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nLCkYY4zxsqRgjDHGy5KCMcYYr/8Pg8U+XzFKnvkAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f9d506d5cd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "N = 8\n",
    "\n",
    "ind = np.arange(N)  # the x locations for the groups\n",
    "width = 0.35       # the width of the bars\n",
    "\n",
    "label = [\"ASW\", \"ARI\", \"NMI\"]\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "for i in [0, 1, 2]:\n",
    "    ax.errorbar(latent_d, np.mean(res[:, :, i], axis=0), yerr=np.std(res[:, :, i], axis=0), label=label[i])\n",
    "\n",
    "# add some text for labels, title and axes ticks\n",
    "ax.set_ylabel('Clustering metrics')\n",
    "ax.set_xticks(latent_d)\n",
    "ax.set_xlabel(\"Dimension of latent variables\")\n",
    "plt.legend()\n",
    "\n",
    "#plt.savefig(\"zeiler_clustering_dimension_stability.pdf\", dpi=300)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEKCAYAAAD9xUlFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzt3Xd8VGX2+PHPSS+kJ9QkNOmioAgi7KprAyuWr6JrQVdx\nVazo2nZddNfu7lr3J651dxUsq4IVC2DDQlFAmgJSAkgJJCE9mTm/P+7NMAkpI2Qyk+S8X695zW1z\n75kh3HOf57n3eURVMcYYYwAiQh2AMcaY8GFJwRhjjI8lBWOMMT6WFIwxxvhYUjDGGONjScEYY4yP\nJQVjjDE+lhSMMcb4WFIwxhjjExXqAH6pzMxM7dGjR6jDMMaYVmXhwoU7VDWrqe1aXVLo0aMHCxYs\nCHUYxhjTqojI+kC2s+ojY4wxPpYUjDHG+FhSMMYY42NJwRhjjI8lBWOMMT6WFIwxxvhYUjDGGONj\nScEYY4yPJQVjAJ47yXkZ085ZUjDGmNaghS5cLCkYY5rfs2Ph6ROgsgQ81aGOxvwCra7vI2NMC/N6\noaIQSvKhNB9KdzjvJe57zct/vrLY+ew9XZ13iYDIWIiKcd4jY/ZM+95jITK6zrJAt6/Zps77Xsvc\nz9QsEwnd7xqmLCkY095UV9R/Iq877z+tnvr3FRUPiZmQkA4JmZBxgDO/8h0nERx6EVRXgqfCOa6n\nyp32X1bpvKoroap0z7qaZf7beyqb97eI9E80MbWThy8hRe+dWALavm5yqy+p1ezDb1lEZPN+x1/I\nkoIxrZkqlBfWc2KvuZqve3WfD5W7G9iZQHwaJGQ4J/aM3pAz3D3pZzgn/YQMSMzYMx+TUP+utixx\n3kdf3/zft97E0liyqaj9Hugy33u58xv7llXVPkZ1RcNJc19IZP2loN1bICWn+Y7TAEsKxoST6kq/\nE3xDJ/YdULpzz7y3gTr7qDj3RJ7unNjTe9V/Yq9JAvFpzXeVevE7zbOfukScq+qoGIgNziH2idcT\nYLJxk0rdZdWVdbavZ9naAogI/inbkoIxwaIKFUWBndhr5isKG95fzVV8Qiak9YDsQ/fM+67mM/ac\n5KMTrM68pUREQkQ8RMcH7xgtdMu0JQVjAuWpqudEXl+1zc4967xV9e8rMrZ2XXxq9/pP7DUn/fg0\niLT/rib47K/MtF+qUF4AhZugbKdz0v/i0YarbcobuYqPS91zIk/Nha5D9z6x16qLT2zTV/EXv38x\nAM+NeS7EkZhfypKCabsqdjsn/KI8933Tnvmizc50VUntz3z4J6dRz7/uPXVo03XxkdGh+Y7GNDNL\nCqZ1qipzT+x5tU/2/if/vernBTp0gpRukNUfDjgWkrtBSjf++tkfKYqM4oELP4eYDm36Kt60UsFq\nvK/DkoLZdzUNX839x1pdCbs3+53g6znxl+3c+3MJmc4JP60n9BjtnvCznffkrpDUxblrpR5r5t/t\nTMQmNe93aYdKq0oprSrFox7WFKwhJTaFlNgUoiOsNNUaWFIwLctTDcU/731l7zvpb4LibYDW/lxc\nCiRnOyf9bsOc95r5ZPcVHReSr9QeFVcWs3H3Rjbs3sCGog2+9427N7K9bLtvu3EzxvmmO0R38CWI\n1NhU33vNdN351NhUOkR3QKzU1qIsKZjm4/VCyfY61Th5tU/4u3/e+0GfmA6+ahw6Ddoz7X+lH9sh\nNN+pHSuqLGJjkXPiX1+03kkCbgLYWV67pJYVn0VOUg6juo0iNymXt9e+TaREcungSymsLKSgooDC\nCue9oKKAwvJCNu7eSEFFAbsbfJgOIiWy3kSSEpNCalw9icVdHhsZTg8xNI+Wary3pGD2zY4foXAj\nVJY6nZ8VbXLq+OveghkZu+cE3/PXfid8v6v8uBSrww8BVaWworD2SX/3Bl8iKKgoqLV9p4RO5Cbn\ncnTO0eQk5dA9uTs5STnkJOWQEF37yeZ5m+cBcGKvE5uMo9pbTVFlkS9x1CSPWknEXb65eDPL85dT\nWFFIhaeiwX3GR8XXSiQpMXVKJ3FOIkmOSfYllaSYJCJD3MVEOAhqUhCRMcAjQCTwtKreV2f9P4Cj\n3dkEoKOqpgYzJrMf8tfAsjdg2ZuwdamzLDIWUMg+bO+TfUq2c4eOnfBDRlXZWb6Tjbs3sr5ofa2T\n/obdG2pdpQtCl8Qu5CTncFz348hNyiU3OZfcpFyyk7KJiwpO9VxURBTpcemkx6X/os+VVZfVSiJ1\nk4p/YtlastVZV1mIV7317k8QkmOTfSWRQKu54qPi21QVV9CSgohEAk8AxwF5wHwRmamqy2u2UdXr\n/ba/GhgarHjMPtq1zk0Eb8CWxc6y7OFwwr3w/etOvywtdFeEqZ+qsqNsR62r/Zr6/Q27N1Did9tt\nhETQNbErucm5nJh54p4Tf3Iu2R2yiYmsvyE+HMVHxRMfFU/nxM4Bf8arXnZX7m4ykRRWFLKjbAdr\nCtZQUFFAaXVpg/uMiYipN1nU10aSGptKcmxyWDe8B7OkMBxYraprAURkOnAasLyB7c8F/hzEeEyg\nCjbuSQSbFznLuh0Kx/8VBo6DVLdTrpWWDFqKV71sK922p1HX74p/4+6NlFWX+baNkii6JXUjJymH\noR2Hkpuc66vu6ZrYlegWeKYiXB9ai5AI3wk7l9yAP1fpqWy8aquykIJyZ/6nwp98y6u14bEkahre\n/ZNFfUmkJt5qbzWREvzqrWAmhW7ARr/5PGBEfRuKSHegJzA7iPGYxhRuguUzYNnrkDffWdZlCBx7\nJwwa5/S1Y4LK4/WwtXTrXo26G3dvZOPujbXq0KMjoslOyiY3KZfhnYeTm5xL96Tu5CTn0CWxC1Et\n0HFaexITGUNWQhZZCVkBf0ZVKa0urdW4XlBR0GDDe007TmMN7zlJ7aeX1PHAa6r19z8rIhOBiQC5\nuYFnd9OE3T+7ieAN2PCls6zTYDjmDqdEkNE7tPG1QdXearaUbNnrNs4NuzeQtzuPKr+G+tjIWF9D\n7qiuo3zVPLlJuXRK6GSNomFOREiMTiQxOpFuHboF/Dn/hveiiiJf4pi6eCpJ0cF/jiaYSWET4J/W\nst1l9RkPXNXQjlT1KeApgGHDhmlD25kAFG9zE8GbsP4LQKHjQDj6j06JILNP4PtqS20JnQc3266q\nvFVs2r3Jd5Xvf8W/afemWlUK8VHx5CTl0DulN0flHEVuUq7vrp6OCR2JEBsxt71pqOF9xuoZLXP8\nIO57PtBHRHriJIPxwHl1NxKR/kAa8GUQY2nfSvJhxUynamjd56BeyOwLR94Mg06Hjv1DHWGrU+mp\nJG93Xq2Ht2oSwJaSLXj8Cr0JUQl0T+5Ov7R+e93Vkxmf2abuXDGtX9CSgqpWi8gkYBbOLanPquoy\nEbkLWKCqM91NxwPTVdVKAM2pdCesfNupGlr7ifPAWHpv+NVkNxEMtFtFm1BeXb7Xvfs101tKtqB+\nT10nRSeRm5zL4MzBnNjrxFpX/Olx6XbiN61GUNsUVPVd4N06y+6oMz8lmDG0K2UFzh1By96AtXOc\nEbnSesCoa51E0HlwsyaCttA9sqqSV5zHrvJdlHvKmTJviu/qf2vp1lrbpsamkpuUy9BOQzkt6TTf\nHT25SbmkxKbYid+0CeHS0Gx+Cf+O6MqLYNV7TtXQ6o+dJ4pTcuHwK+HAM5w7iOxkBTh396wrWsfy\n/OWs3LmSFTtXsDJ/Jbur9tztMWfjHHKTchnRZQQ5STm+K/7spGxSYlNCGL0xLcOSQmvk9Ti9hE7/\nLfz4oTOea3I3GHE5DDoDuh3S7hNBpaeS1QWrWZG/ghU7ndcPO3+g3FMOOHf29E3ry9ieY+mf0Z9X\nV71KXGQc/z7x3yGO3Jj6tVSJ3JJCa/Pz97BlkTOQd0URDLvEKRF0GwYR7fNOldKqUn7Y9QPL85c7\nV/87V7K6YDXV7oD2HaI70C+9H2f1PYsBGQMYkD6Anik9a93L/+7adxvavTHtiiWF1uT712HGVc4w\nkh0Hwe8/b3eJoLCi0Ffts3ynUw20rnCdr9E3LTaNARkDuGjgRb4EkJ2Ubbd2GhMgSwqtgdcDH98F\nXzwMOSOcpBAZ0+YTwvbS7U7Vj1sFtHLnSjYV73nUpXNiZ/qn92dsj7EMyBhA//T+dEroZA2+xuwH\nSwrhrmwXvPY7WPOxU1U05n74z+mhjqpZqSqbijfVSgAr8leQX57v26Z7cncGZw7m//r+n68EkBaX\nFsKojWmbLCmEs63LYfp5zkA1pzwCh05wlrfiJ4k9Xg/ri9Y7VT/5K32NwDX9vURKJL1SezGq2ygG\npA9gQMYA+qX1o0OMDbJjTEuwpBCuls+AN65wRhyb8A7k1tuXYFir8lQ5dwD5lQB+2PWDr0fPmIgY\n+qb1ZUyPMfRP78/AjIEckHpA0PrtN8Y0zZJCuPF6YM498NlDzsA1Z/8HkruEOqom1dwBVJMAVu5c\nyY8FP/ruAEqMTqRfWj/O7HMm/dP7MyDDuQMoXPuUN6a9sqQQTsoK4PXL4McP4JAL4cSHnEFswlC1\nt5odZTsoqizitDdPY13ROt+IVmmxafRP78+FAy/01f/nJOXYHUDGtAKWFMLFtpVO+0HBejjp706j\ncpjeRTNv8zwenP8g64rWER0RzUGZB3FCjxN8bQB2B5AxrZclhXCw4i144/cQnQAXvQ3dR4Y6onqt\nL1rPQwseYu7GuWR3yKZ3Sm9SY1N57JjHQh2aMaaZWHk+lLxemH03vHw+ZPWDyz8Jy4Swu3I3f1vw\nN8bNGMc3W77h+kOvZ8a4GaTFpVmJwJg2xkoKoVJeCK9PhB/eh6Hnw4l/g+jwuuvG4/Xw5uo3efTb\nR9lVvotxB4zjmkOuITM+M9ShGWOCxJJCKGz/wWk/2PWT05h82KVh136w4OcF3D//flbuXMnQjkP5\n57H/ZFDGoFCHZYwJMksKLW3lu04JIToOLpwJPUaFOqJaNhVv4u8L/s4H6z+gc2JnHvz1g5zQ44Q2\nX03UmseEMKY5WVJoKV4vfPoAzL0Xug6Fc/4LKdmhjsqntKqUZ75/hue/f54IieDKIVcyYdAE4qPi\nQx2aMaYFWVJoCeVFzt1Fq96Bg8+Dk/8O0eFxsvWql3fWvsPDCx9mW9k2Tup1Etcdch2dEzuHOjRj\nTAhYUgi2Hath+rmQv8bpzG7E5WHTfrB4+2Ie+OYBluxYwoEZB/K3o/7GkI5DQh2WMSaELCkE0w+z\n4H+XQmQ0XDgDev4q1BEBsLVkKw8vepi3175NVnwWd4++m5N7nWxPHBtjLCkEhdcLn/0N5twNnQfD\n+BchNTfUUVFeXc7zy57n2e+fxeP1cNngy7h08KUkRCeEOjRjTJgIalIQkTHAI0Ak8LSq3lfPNmcD\nUwAFFqvqecGMKegqdsObVzhPKR90jtPldYjbD1SVWetn8Y8F/2BzyWaO634cNxx6A9lJ4dPQbYwJ\nD0FLCiISCTwBHAfkAfNFZKaqLvfbpg9wKzBKVXeJSMdgxdMi8tc4zx/s+BFOuAcOvzLk7QfL85dz\n/zf3s2jbIvql9ePZ0c9yWOfDmmXfdhunMW1PMEsKw4HVqroWQESmA6cBy/22uQx4QlV3AajqtiDG\nE1w/fuiMkBYRCRe8Dr2OCmk4O8p28Ni3j/HGj2+QFpfGHSPv4IwDziAyIjKkcRljwlswk0I3YKPf\nfB5Qd6SYvgAi8gVOFdMUVX0/iDE1P1X4/O/w8V+g04FO+0Fa96Ae8uL3Lwbqv1Kv9FTy4ooXmbpk\nKhXVFVw48EIuP/hykmKSghqTMaZtCHVDcxTQBzgKyAY+FZHBqlrgv5GITAQmAuTmhr7B1qeiGGZc\n6YySduCZcOrjEBOaRltVZc7GOTy04CE27t7IkdlHcuOwG+mR0iMk8RhjWqdgJoVNQI7ffLa7zF8e\n8LWqVgE/icgPOElivv9GqvoU8BTAsGHDNGgR/xI718L038L2lXDcX+CIq0PWfvDjrh95YP4DfLXl\nK3ql9OLJY59kVLfw6j7DGNM6BDMpzAf6iEhPnGQwHqh7Z9GbwLnAcyKSiVOdtDaIMTWP1R857QcA\n5/8Pev8mJGEUlBfw+HeP8+oPr9IhugO3DL+Fs/udbUNcGmP2WdCSgqpWi8gkYBZOe8GzqrpMRO4C\nFqjqTHfd8SKyHPAAN6lqfrBi2m+q8MUj8PGdkDXAaT9I79niYXjVy/ay7Zz0xkmUVJVwdt+zuWrI\nVaTGpbZ4LMaYtuUXJQURiQA6qGpRINur6rvAu3WW3eE3rcAN7iu8VZbAjEmw7HUYdDqc9gTEJLZ4\nGMWVxazatYqSqhJGdhnJHw77AwekHdDicRhj2qYm+zUQkZdEJFlEEoHvgeUiclPwQwsju9bBM8fD\nsjfg2Clw1nMhSwi//+j3lFaV0jO5J1OPm2oJwRjTrALp7GagWzIYB7wH9AQuCGpU4WTNHHjqKCjc\nCOe/BqOvD0mDcklVCVd8dAXLdiyjV0ovMuIz2vwYB8aYlhdIUogWkWicpDDTvVMoPO4ACiZVmPcY\n/PcM6NAZLpsDBxwbklBqEsLSHUt54MgHSItLC0kcxpi2L5CkMBVYByTiPEfQHQioTaHVqiyF1y+D\nD/4I/U+GSz+CjN4hCaWkqoQrP7qSJduX8OCRD3Jc9+NCEocxpn1osqFZVR8FHq2ZF5ENwNHBDCqk\ndq2Hl38LP38Pv/kT/GpyyJ4/KK0q5cqPrnTGPfj1A5YQjDFB12RSEJF7gAf8njJOBSYDfwxmYCGx\n9hN4dQJ4PXDeK9D3+JCFUlpVyhUfXcHi7Yu5/9f3c3yPPbFYR3TGmGAJpPporH+3E27ndScGL6QQ\nUIUv/wn/OR0Ss2DinJAnhCs/dkoI9/36Pk7ocULIYjHGtC+BPKcQKSKxqloBICLxQGxww2pBVWXw\n1rWw5GWn/eD0JyE2dJ3H1SSEb7d9y/2/up8xPcaELBZjTPsTSFJ4EfhYRGrqLC4GXgheSC2oYKPT\nfrBlMRx9O/zqRogI3ZCUpVWlXPXxVXy77Vvu+9V9jOlpCcEY07ICaWi+X0SWAMe4i/6iqrOCG1YQ\nPHeS837xO877us/hlYvAUwnnTod+Y0MXG05CmDR7Eou2LeLe0fcytmdo4zHGtE8BdXOhqu/hPLjW\n+qnCN0/B+7dCei84dxpk9glpSGXVZVw9+2oWbl3IPaPv4cRebavJxhjTejSYFETkc1UdLSK7qf2w\nmuB0W5Qc9Oiam3rhzSth8UvQdyycMRXiUkIaUll1GZM+nsSCrQu4e/TdnNTrpJDGY4xp3xpMCqo6\n2n1vG0N2VVfA9hWwYR4ceQsceXNI2w/ALSF8fDULti7gr6P+ysm9Tg5pPMYY02j1kYhEAstUtX8L\nxRM8JducO43OeREGhP7kW1Nl9M3P33D36Ls5pfcpoQ7JGGMaTwqq6hGRVSKSq6obWiqooEjOhoSs\nsEgI5dXlXDP7Gr7ZYgnBGBNeAmloTgOWicg3QEnNQlU9NWhRBYMIRMeFOgrKq8u5evbVfL3la/46\n+q+WEIwxYSWQpPCnoEfRTtSUEL7e8jV/GfUXTu3duvKqMabtCyQpnKiqN/svEJH7gU+CE1LbVF5d\nzrVzruWrLV9x16i7OO2A00IdkjHG7CWQ22/q65qz9T1ZdfE7ex5ca2EVngqum3MdX27+kjuPuJNx\nB4wLSRzGGNOUxp5TuAK4EujlPtFcIwmYF+zA2ooKTwXXzr6WeZvncecRd3J6n9NDHZIxxjSoseqj\nl3CeYr4XuMVv+W5V3RnUqNqICk8F1865li82f8FdR9xlCcGYEKqqqiIvL4/y8vJQhxJUcXFxZGdn\nEx0dvU+fb+zhtUKgEDhXREYDfVT1ORHJFJGeqvpTUzsXkTHAI0Ak8LSq3ldn/QTgQWCTu+hxVX16\nn75JmKmpMvpi0xdWQjAmDOTl5ZGUlESPHj3a7Pjmqkp+fj55eXn07Nlzn/bRZJuCiPwZuBm41V0U\nA/w3gM9FAk/gtD8MxEkuA+vZ9GVVHeK+2kRCqPRUcv2c6/l80+dMGTmFM/qcEeqQjGn3ysvLycjI\naLMJAUBEyMjI2K/SUCANzacDp+I+o6Cqm3HaFZoyHFitqmtVtRKYDrT5W24qPZVcP/d6Ptv0GX8e\n+WfO7HtmqEMyxrjackKosb/fMZCkUKmqitspnogkBrjvbsBGv/k8d1ldZ4rIEhF5TURyAtx3WKr0\nVHLD3Bv4NO9T7hh5B2f1PSvUIRlj9sM5U7/knKlfNus+33zzTUSElStXAuD1ernmmms48MADGTx4\nMIcddhg//fQTjzzyCNddd53vc5dffjnHHnusb/6xxx7jmmuuadbYILCk8IqITAVSReQy4CPgX810\n/LeAHqp6EPAhDQzeIyITRWSBiCzYvn17Mx26eVV6Kpk8dzKf5H3Cnw7/E//X9/9CHZIxJgxNmzaN\n0aNHM23aNABefvllNm/ezJIlS1i6dClvvPEGqampjBo1innz9tzouXjxYgoLC/F4PADMmzePI444\notnjazIpqOpDwGvA/4B+wB2q+lgA+94E+F/5Z7OnQblm3/k1w3wCTwOHNhDDU6o6TFWHZWVlBXDo\nllXlqWLy3MnMzZvLnw7/E2f3OzvUIRljwlBxcTGff/45zzzzDNOnTwdgy5YtdOnShQi31+bs7GzS\n0tIYMmQIP/zwA2VlZRQWFhIfH8+QIUNYunQp4CSFUaNGNXuMgQ6y86GIfF2zvYikB3Bb6nygj4j0\nxEkG44Hz/DcQkS6qusWdPRVY8UuCDwdVnipu+OQG5ubN5fYRt1tCMMY0aMaMGYwZM4a+ffuSkZHB\nwoULOfvssxk9ejSfffYZxxxzDOeffz5Dhw4lKiqKoUOHMn/+fMrKyhgxYgR9+vRh3rx5ZGVloark\n5DR/jXuTSUFELgfuBMoBL+4gO0Cvxj6nqtUiMgmYhXNL6rOqukxE7gIWqOpM4BoRORWoBnYCE/bj\nu7S4Kk8Vkz+ZzNyNc7ltxG2M7z8+1CEZYwJw51vLWL65qMntlm9xtgmkXWFg12T+fMqgRreZNm0a\n1157LQDjx49n2rRpPPTQQ6xatYrZs2cze/ZsjjnmGF599VWOOeYYjjjiCObNm0dZWRkjR46kT58+\n3HPPPWRlZQWl6ggCKyncCByoqjt+6c5V9V3g3TrL7vCbvpU9t7q2KlWeKm785EbmbJzDbSNu49z+\n54Y6JGNMGNu5cyezZ89m6dKliAgejwcR4cEHHyQ2NpaxY8cyduxYOnXqxJtvvskxxxzDqFGjePLJ\nJykvL+eqq64iKyuL5cuXhzwprAFKg3L0VqrKW8VNn97E7I2zuXX4rZYQjGllmrqir1FTQnj58pH7\nfczXXnuNCy64gKlTp/qWHXnkkXz22WcccMABdO3aFa/Xy5IlSzjooIMAGDlyJBMmTKBbt2507NgR\ngKysLGbMmMGrr7663zHVJ5CkcCswz21TqGkURlWb/16oVqDKW8VNn9zExxs+5pbht3DegPOa/pAx\npt2bNm0aN99cq8NpzjzzTC666CLS09OpqHBOr8OHD2fSpEkApKWlkZWVxaBBe5LYyJEj+eKLLzj4\n4IODEqc4jyA0soEzuM7nwFKcNgUAVLXe20eDbdiwYbpgwYJQHJoqbxV/+OQPfLThI24Zfgu/HfDb\nkMRhjPnlVqxYwYABA37RZ5qzpNCS6vuuIrJQVYc19dlASgrRqnrDvgbXVlR5q7j505v5aMNH3HzY\nzZYQjDFtUiBJ4T0RmYjzoJl/9VG76Sm1JiF8uP5D/nDYHzh/4PmhDskY0wJaWwmhOQSSFGpaUf3v\nEmryltS2ospbxS2f3sKH6z/kpmE3ccHAC0IdkjHGBE2TSUFV963/1Tag2lvNLZ/ewgfrP+DGYTdy\n4aALQx2SMcYEVSB9H7VL1d5qbvlsT0K4aNBFoQ7JGGOCzpJCPaq91dz62a3MWjeLyYdOtoRgjGk3\nLCnUUe2t5rbPbuP9de9zw6E3MOHACaEOyRgTKs+d5LyaUd2us9etW+fr7G7gwIFceOGFVFVVATB3\n7lxOPvnkZj1+UwIZee2Qel69RSSgzvRak2pvNbd9fhvvrXuP6w+9nosPvDjUIRlj2pi6XWcD9O7d\nm++++46lS5eSl5fHK6+8ErL4Aikp/BP4CngKZxyFL4FXgVUicnwQY2tR1d5qbv/8dt776T2uO+Q6\nLjnwklCHZIxpY+rrOttfZGQkw4cPZ9OmTfV8umUEkhQ2A0Pd8QwOBYYCa4HjgAeCGVxL8Xg93P75\n7bz707tce8i1/G7w70IdkjGmDaqv62x/5eXlfP3114wZMyZEEQb2nEJfVV1WM6Oqy0Wkv6qubU3j\nnV78vlMV9NyY52ot93g9/PGLP/oSwqWDLw1FeMaYlvTeLfDz0qa3+3mJ8x5Iu0LnwTD2vkY3qa/r\n7EmTJrFmzRqGDBnCTz/9xEknneTrEC8UAkkKy0Tk/wE1ZZ1zgOUiEgtUBS2yFuDxevjTF3/i7bVv\nc83QaywhGGOCpqGus6+66ipfm8KOHTsYNWoUM2fO5NRTTw1JnIEkhQnAlUDNCNJf4IyxUAUcHZyw\ngs/j9XDHvDt4a+1bXD30ai476LJQh2SMaSlNXNH71JQQLn5nvw/ZUNfZGzdu9M1nZmZy3333ce+9\n94YsKQQyRnOZqv5NVU93Xw+paqmqelW1uCWCbG41CWHmmplMGjKJiQdNDHVIxpg2btq0aZx++um1\nlp155pnce++9tZaNGzeO0tJSPvvss5YMzyeQ4ThHAVOA7v7bq2qr7PvI4/Xw53l/ZuaamVw55Eou\nP/jyUIdkjGkH5syZs9eya665hmuuqT00jYiwePFi3/xRRx0V7NBqCaT66BngemAh4AluOMGlqkz5\ncgoz1szgyoOv5IqDrwh1SMYYE1YCSQqFqvpe0CMJMlVlXdE6Fm5byBUHX8EVQywhGGOa0AxtCa1N\nIElhjog8CLxO7fEUFgUtqiDYUrKF/PJ8fn/w77lyyJWhDscYY8JSIElhhPvuP4ybAr9p/nCCJys+\ni6iIKK482BKCMcY0JJDxFPb5tlMRGQM8AkQCT6tqvfeBiciZwGvAYaoalAGYoyOj6ZjQkdb0wJ0x\nxrS0BpM4CDq5AAAc4klEQVSCiJyvqv8VkXrHZ1bVvze2YxGJBJ7A6Q4jD5gvIjNVdXmd7ZKAa4Gv\nf2nwxhhjmldjzykkuu9JDbyaMhxYraprVbUS54no0+rZ7i/A/UB5oEEbY0xLuPj9i31d5DQHEWHy\n5Mm++YceeogpU6YAMGXKFESE1atX+9Y//PDDiAgLFjgVKD169GDHjh3NFk99GkwKqjrVvdovUtU7\n674C2Hc3YKPffJ67zEdEDgFyVLXRJn4RmSgiC0Rkwfbt2wM4tDHGhJ/Y2Fhef/31Bk/sgwcPrtV7\n6quvvsqgQYNaKjygiSeaVdUDnBuMA4tIBPB3YHJT26rqU24vrcOysrKCEY4xxgRdVFQUEydO5B//\n+Ee968eNG8eMGTMAWLNmDSkpKWRmZrZkiAF1nf2FiDwuIr/yH2gngM9tAnL85rPdZTWSgAOBuSKy\nDjgcmCki/nc5GWNMm3LVVVfx4osvUlhYuNe65ORkcnJy+P7775k+fTrnnHNOi8cXyC2pQ9z3u/yW\nBXJL6nygj4j0xEkG44HzfDtQLQR8KVBE5gI3Buvuo7pdZhtj2q/7v7mflTtXNrldzTaBtCv0T+/P\nzcNvbnK75ORkLrzwQh599FHi4+P3Wj9+/HimT5/OrFmz+Pjjj3nuuZY9dwXtllRVrRaRScAsnFtS\nn1XVZSJyF7BAVWfuy36NMaa1u+666zjkkEO4+OK9k83JJ5/MTTfdxLBhw0hOTm7x2ALpEK8TcA/Q\nVVXHishAYKSqPtPUZ1X1XeDdOsvuaGDbowKK2Bhj9lMgV/TQ8OBc+ys9PZ2zzz6bZ555hksuqT30\nb0JCAvfffz99+/Zt1mMGKpA2hedxrva7uvM/sGdsBWOMMftg8uTJDd6FNH78eA45JJCm2+YXSJtC\npqq+IiK3gq9aqFX3lmqMMaFQXLxnCJpOnTpRWlrqm695XqGuuXPn+qbXrVsXpMj2CKSkUCIiGTiN\ny4jI4cDezebGGGNavUBKCjcAM4HeIvIFkAX8X1CjMsaYMNAe71oMJCksA44E+gECrCKwEoYxxphW\nJpCT+5eqWq2qy1T1e1WtAr4MdmDGGNPcVDXUIQTd/n7HxnpJ7YzTV1G8iAzFKSUAJAMJ+3VUY4xp\nYXFxceTn55ORkdFmu9BXVfLz84mLi9vnfTRWfXQCMAGne4q/sScp7AZu2+cjGmNMCGRnZ5OXl0db\n71QzLi6O7Ozsff58g0lBVV8AXhCRM1X1f/t8BGOMCQPR0dH07Nkz1GGEvUDaFLJFJFkcT4vIIhE5\nPuiRGWOMaXGBJIVLVLUIOB7IAC4A6h1W0xhjTOsWSFKoaUs4Efi3qi7zW2aMMaYNCSQpLBSRD3CS\nwix3TGVvcMMyxhgTCoE8vPY7nDEV1qpqqdvlRfMNWmqMMSZsBJIURrvvB7XVe3uNMcY4AkkKN/lN\nxwHDgYU0PfKaMcaYViaQkddO8Z8XkRzg4aBFZIwxJmT2pWO7PGBAcwdiAnfO1C85Z6p1P2WMaX6B\nDMf5GO5YCjhJZAiwKJhBGWOMCY1A2hQW+E1XA9NU9YsgxWMCoKp4vEpReRWRIkRGCCL4TbfMDQE1\npZWXLx/ZIsczxgRfIG0KL7REICYwhWVVLNtSREmFh4OmfNDgdpERQqS4yaLutJs4apJIRAREuPMR\nEUKEuPMRQoS7LNJdFhGx53Mrfy4iOjKCx2f/SM/MDvTITKBnZiIJMYFcaxhjwlFjXWcvZU+1Ua1V\ngKrqQU3tXETGAI8AkcDTqnpfnfW/B64CPEAxMFFVlwcefvtSXFHNhOe+obTCQ7fUeC4e1QOvKl4F\nj1fxehWPO++b9ipeVTxe3G2dUsaed/b+nDtfUyLx6J7SicerVHu9VHqg2quUVlbx0Ac/1Iqzc3Kc\nmyA60CszkZ6ZifTITCQ3PYGYKBufyZhw1tgl3cn7s2MRiQSeAI7DaZyeLyIz65z0X1LVJ93tTwX+\nDozZn+O2VaWV1Vzy3HyW5BVyQMcOpCfGcOmveoU0pprqo+cuPox1O0r5aUcJP+0o5qcdpfy0o5j3\nv9/CrtIq3/YRAjnpCfTIcBJFryw3YWQk0jU1nsgIew7GmFBrLClEA53qth+IyCjg5wD2PRxYrapr\n3c9NB04DfEnB7WivRiL1l0zavfIqD5f9ewEL1u/kkfFD+e9X60MdElC7LWFg12QGdk3ea5uC0ko3\nWdR+zV+3k9JKj2+7mKgIemS4CSMr0S1hOFVSWR1i2+ygKMaEm8aSwsPArfUsL3LXnVLPOn/dgI1+\n83nAiLobichVwA1ADPZA3F4qqj1c8d+FzFuTz0NnHcwpB3fllIO7hjqsgKUmxDA0N4ahuWm1lqsq\n23ZX7JUs1u4oYc6qbVR59lwfdIiN8lVB9cx0EkbNdEp8dLPEaY3mxjgaSwqdVHVp3YWqulREejRX\nAKr6BPCEiJwH/BG4qO42IjIRmAiQm5vbXIcOe1UeL1e/9C1zVm3nntMHc+ah+z6aUrgRETolx9Ep\nOY7De2XUWufxKpt2lfFTfgk/bS92EkZ+Kd9t3MXbSzbjPwRtRmJMvQmjR0Yi8TGRLfytjGn9GksK\nqY2siw9g35uAHL/5bHdZQ6YD/6++Far6FPAUwLBhw9pFFVO1x8t1L3/HB8u3MuWUgZw3ov0kw8gI\nITcjgdyMBI7sm1VrXUW1h407S1m7vXYJ49MftvPawrxa23ZNiaNnVmKtNoweGYnkpCcQHWkN3sbU\np7GksEBELlPVf/kvFJFLcfo+asp8oI+I9MRJBuOB8+rsq4+q/ujOngT8iMHrVf7w2hLeWbKF207s\nz4RRNoRgjdioSA7omMQBHZP2WldcUc26etov3lq8maLyat92kRFCbnoCPTKcO6R6ZiVSWFZFdKSw\no7iCtIQYa/Q27Zao1n/hLSKdgDeASvYkgWE4df+nq2qTjc0iciJO+0Mk8Kyq3i0idwELVHWmiDwC\nHAtUAbuASe4gPg0aNmyYLliwoLFNWjVV5bY3ljLtm41MPq4vVx/TJ9QhtXqqyq7Sqlp3RjkJw5ku\nr6o9PIgIpCXEkJEYQ3piDJkdYklPjCGjg7Msw53P7BBDemIsqfHRRFgSMWFORBaq6rAmt2soKfjt\n6GjgQHd2marObob49llbTgqqyp1vLef5eeuYdPQB3HhCv1CH1OZ5vcrW3eVc8tx8qjxeLhjZg/zi\nCvJLKskvrmRnSSU7SirYWVJJgd/ttf4iBCdpJDaePGqWJ8dZEjEtL9CkEMgTzXOAOc0SlWmQqnLv\neyt5ft46LvtVTyYf3zfUIbULERFCl5R4kt27mC46okeD21Z5vOwq9UsWxU6yyC+udJOIM79scxH5\nxRW1qqz8RUUIaYk1iWNPMqmbPGqSSnJclN2Sa1qM9UcQJv7x4Q889elaLhzZndtOHGAngTAUHRlB\nx6Q4OibFBbR9ZbWTRPyTx16JpKSCvF0F5BdXUlxRfxKJjhRfSaQmYdSXPJykEkOHWEsiZt9ZUggD\nj8/+kUdnr2b8YTlMOWWQ/YduI2KiIny33QaivMrDzpLGSyE7SipZl19CfnFlrYf/6h63phSSnhhL\npts2ktEh1m/5nraShJhI+5szPpYUQuxfn67loQ9+4Iyh3bj79MFW1xwi4fDQWlx0JF1T4+maGsgd\n31BW6SG/pIHkUVzJzhKnbWTNtmLySyr2alDfc9wIXylk7xJJ3Yb2WHv+o42zpBBC//5yHXe/u4KT\nDurCA2cdZLdBml8kPiaS7JgEstMSAtq+tLK6VvKoaUz3L4XsKK5g1c+7yS+ppLK6/iSSEBNZu+SR\nGEN6hxgy/RraH5q1iujICP53xRF2odPKWFIIkenfbOCOGcs4bmAnHj5nCFH2MJUJsoSYKBLSo8hJ\nbzqJqCrFFdV+pY7KOndlOdM/F5azbHMhO0sqa3VNUqPfn96jS0o8XVPj6JaaQLe0eLq5011T4+ia\nGk9ctJU8woklhRB4fVEet76xlKP6ZfH4eUPt6VoTdkSEpLhokuKi6Z6R2OT2qkpRebUvedz6+lKq\nPF5OOLAzmwvK2bSrlC9W72Dr7nLq3gWf2SGGbm61WbfUeLql+U2nxpOaEG1tHi3IkkILe3vJZm58\ndTFH9M7gyfMPJTbKrpJM6ycipMRHkxIfTc/MRNITYwC4dWzt4dwrq71sLSonb1cZmwvK2FRQxqZd\nZWwuLGPVz7uZvXIbFXWqrRJiIn1JomtqPNlptUsenZJiraTdjCwptKBZy37m2unfMax7Ov+6cJgV\nm02b1VDDfUxUBDnpCQ1WYakq+SWVTsLY5SaNgj0JZOkmp6rKX2SE0Dk5zk0UbinDLW1ku/OJsXaq\nC5T9Ui1kzsptTHppEQdlp/DsxYfZkJXG1ENEyOwQS2aHWA7Krr9PztLKajdJlDuljII9yWP+ul38\nXLQFj7d2HVVqQnTtKqo61VSZHWKsisplZ6YW8MXqHVz+34X065zE8xcPp4NdtRizzxJiohrsFBGc\nrte3FpX7Shj+VVXr80uYt3oHJXWe8YiJivAli65+DeFOw3g8XVLi281QsnZ2CrKv1+bzuxfm0ysz\nkf9cMqLZBoUxxtQvMkIafd5DVSkqqyavoLRWSWNzQTl5BWXMWbWd7bsran1GBLI6xPqShK+kkbKn\nxNFW/m9bUgiiRRt2ccnz8+mWGs9/Lx1Bmtv4ZowJHREhJSGalIQUBnVNqXeb8ioPPxeW+6qlato3\nNrvtGh8s20qlp3aDeFJslF97xt634GYlxbaKZ5EsKQTJ0rxCLnr2G7KSYnnpssPJ7BAb6pCMMQGK\ni450RvDLrP92XK9X2VFcUbshfJfbzlFQxsL1uygsq92rblSE0CU1zle66OZ3R1XNfDjcfGJJIQhW\nbCnigme/JiU+mpcuOzzgvm+MMa1DRITQMTmOjslxe40/XmN3eRWbC8qddg2/xLG5oIwv1+Sztaic\nOu3hZCTG7FUt5V9dldYCz2xYUmhmq7ft5vynvyY+OpJplx0ecD82xpi2JSkumn6do+nXuf4G8SqP\n11dFtdnveY28XWX8uG03c3/Ytld/Vd3TE/jkD0cHNW5LCs3opx0lnPevr4mIEF68dERA3QkYY9qn\n6Mimn9nYVVrla8+4590VJMcH/5RtSaGZbNxZynn/+opqr/LyxMPpldUh1CEZY1oxEWccjfTEGAZn\npzDmwM4tctz2ceNtkG0uKOO8p7+itNLDf383gj6d6i8uGmNMuLOksJ+2FZXz26e/pqCkiv/8bjgD\nuyaHOiRjjNlnVn20H3YUV3De01+ztaic//xuRIOP5RtjTGsR1JKCiIwRkVUislpEbqln/Q0islxE\nlojIxyLSPZjxNKeC0krOf/pr8naV8uyEwzi0e/23pRljTGsStKQgIpHAE8BYYCBwrogMrLPZt8Aw\nVT0IeA14IFjxNKei8ioueOYb1u4o4ekLD+PwXhmhDskYY5pFMEsKw4HVqrpWVSuB6cBp/huo6hxV\nLXVnvwKygxhPsyiuqGbCs9+w8ucinjz/EEb3yQx1SMYY02yCmRS6ARv95vPcZQ35HfBeEOPZb2WV\nHi55fj6L8wp57NxD+E3/TqEOyRhjmlVYNDSLyPnAMODIBtZPBCYC5ObmtmBke5RXebjs3wtYsG4n\nj4wf2mL3DBtjTEsKZklhE5DjN5/tLqtFRI4FbgdOVdWKuusBVPUpVR2mqsOysrKCEmxjKqu9XPHf\nhXyxZgcPnnUwpxzctcVjMMaYlhDMpDAf6CMiPUUkBhgPzPTfQESGAlNxEsK2IMayz6o8Xq6etog5\nq7Zz97jBnHlo2Dd7GGPMPgtaUlDVamASMAtYAbyiqstE5C4ROdXd7EGgA/CqiHwnIjMb2F1IeLzK\nDa8sZtayrUw5ZSDnjQhN1ZUxxrSUoLYpqOq7wLt1lt3hN31sMI+/P7xe5Q+vLeGtxZu57cT+TBjV\nM9QhGWNM0Fk3F/VQVW5/83v+tyiPG47ry8Rf9w51SMYY0yIsKdShqtz51nKmfbOBq47uzdW/OSDU\nIRljTIuxpOBHVbnvvZU8P28dl47uyY3H9wv6KEfGGBNOLCn4+cdHPzL107VccHh3bj9pgCUEY0y7\n026SwjlTv+ScqV82uP6JOat59OMfOWdYDneeOsgSgjGmXWo3SaExT3+2lgdnreL0od2454zBRERY\nQjDGtE/tPin8+8t1/PWdFZw0uAsPnnUQkZYQjDHtWLtOCi/P38AdM5Zx3MBOPDx+CFGR7frnMMaY\n9psU3vg2j1teX8qRfbN4/LyhRFtCMMaY9pkU3lmyhcmvLGZkrwymXnAosVGRoQ7JGGPCQrtLCh8s\n+5lrp3/Lod3TePqiYcRFW0Iwxpga7SopFJRWctVLiziwWwrPTjiMhJiwGE7CGGPCRrtJCoVlVfyw\nrZh+nZN44ZLhJMVFhzokY4wJO+0mKVR7lfjoSP5zyQhS4i0hGGNMfdpN/UlGYgzpCdGkJcaEOhRj\njAlb7aakAFjXFcYY04R2lRSMMcY0zpKCMcYYH0sKxhhjfCwpGGOM8bGkYIwxxseSgjHGGJ+gJgUR\nGSMiq0RktYjcUs/6X4vIIhGpFpGzghnLy5eP5OXLRwbzEMYY0+oFLSmISCTwBDAWGAicKyID62y2\nAZgAvBSsOIwxxgQumE80DwdWq+paABGZDpwGLK/ZQFXXueu8QYzDGGNMgIJZfdQN2Og3n+cu+8VE\nZKKILBCRBdu3b2+W4IwxxuytVTQ0q+pTqjpMVYdlZWWFOhxjjGmzgpkUNgE5fvPZ7jJjjDFhKphJ\nYT7QR0R6ikgMMB6YGcTjGWOM2U9BSwqqWg1MAmYBK4BXVHWZiNwlIqcCiMhhIpIH/B8wVUSWBSse\nY4wxTQvqeAqq+i7wbp1ld/hNz8epVjLGGBMGWkVDszHGmJYhqhrqGH4REdkOrN/Hj2cCO5oxnFAK\nl+8SLnE0h7b0XcKB/Z7Nb39+0+6q2uTtm60uKewPEVmgqsNCHUdzCJfvEi5xNIe29F3Cgf2eza8l\nflOrPjLGGONjScEYY4xPe0sKT4U6gGYULt8lXOJoDm3pu4QD+z2bX9B/03bVpmCMMaZx7a2kYIwx\nphHtIimISI6IzBGR5SKyTESuDXVM+0NE1onIUhH5TkQWhOD4/dxj17yKROS6lo5jX4nIsyKyTUS+\n91uWLiIfisiP7ntaKGNsbRr4TaeIyCa/v5MTQxlja9LQOasl/k7bRfWRiHQBuqjqIhFJAhYC41R1\neRMfDUsisg4YpqohvwfcHUxpEzBCVff1+ZEWJSK/BoqBf6vqge6yB4CdqnqfO0pgmqreHMo4W5MG\nftMpQLGqPhTK2Fqjhs5ZOIOSBfXvtF2UFFR1i6oucqd34/TFtE9jO5i9HAOsaS0JAUBVPwV21ll8\nGvCCO/0Czn9AE6AGflOzjxo5ZwX977RdJAV/ItIDGAp8HdpI9osCH4jIQhGZGOJYxgPTQhxDc+ik\nqlvc6Z+BTqEMpg2ZJCJL3Oolq5LbB3XOWUH/O21XSUFEOgD/A65T1aJQx7MfRqvqITjjX1/lFt1b\nnNsl+qnAq6E4frCoU6fa9utVg+//Ab2BIcAW4G+hDaf1aeycFay/03aTFEQkGufHfVFVXw91PPtD\nVTe579uAN3DGww6FscAiVd0aouM3p61uPW5Nfe62EMfT6qnqVlX1qKoX+Beh+zttlRo4ZwX977Rd\nJAUREeAZYIWq/j3U8ewPEUl0G54QkUTgeOD7xj8VNOfSNqqOwBkA6iJ3+iJgRghjaRNqTl6u0wnd\n32mr08g5K+h/p+3l7qPRwGfAUsDrLr7NHe+hVRGRXjilA3DGw3hJVe8OQRyJwAagl6oWtvTx94eI\nTAOOwulxcivwZ+BN4BUgF6cX3rNV1RpOA9TAb3oUTtWRAuuAy/3qw00jGjpn4bQrBPXvtF0kBWOM\nMYFpF9VHxhhjAmNJwRhjjI8lBWOMMT6WFIwxxvhYUjDGGONjScEETEQ8bm+Xy0RksYhMFpEId90w\nEXk0RHHNC/L++7vf+1sR6V1n3ToRyWzi87ft5/HHicjA/dlHI/s+1e1YrbFtpojIjfUs7+HfK6pp\nGywpmF+iTFWHqOog4DicJ5r/DKCqC1T1mlAEpapHBPkQ44DXVHWoqq7Zh8/vV1Jwj9/sSUFEolR1\npqre19z7Nq2XJQWzT9wuNibidHgmInKUiLwNvivLF0TkMxFZLyJniMgD7hgQ77uP7yMih4rIJ27H\nfrP8Ht+fKyL3i8g3IvKDiPzKXT7IXfad28laH3d5sfsuIvKgiHzvHuscd/lR7j5fE5GVIvKi+8Ro\nLSIyRES+cvf9hoikiTMGwHXAFSIyp7HfRETedL/LspqOCkXkPiDejflFd9n5ft9jqjjdjyMixSJy\nt1sK+0pEOonIETj9Sz3obt/b73gp7u9bU1pLFJGNIhItIpeJyHx3X/8TkQR3m+dF5EkR+Rp4QEQm\niMjj7rpTRORrt0T0kYj4d7Z2sIh8KU4//pfV890j3d9+vvv7Xe4u7yIin7qxf1/zb2nCmKray14B\nvXD6xq+7rACnp8ajgLfdZVOAz4Fo4GCgFBjrrnsD58o3GpgHZLnLzwGedafnAn9zp08EPnKnHwN+\n607HAPH+cQFnAh8CkW5MG4AubmyFQDbOhdCXOJ0K1v0uS4Aj3em7gIf9vs+NDfwm64BMdzrdfY/H\n6dIho+7vBgwA3gKi3fl/Ahe60wqc4k4/APzRnX4eOKuB488Ajvb7DZ92pzP8tvkrcLXfvt4GIt35\nCcDj7nQaex5ovdTv32AKsNj9XpnARqAr0AP43t1mol+8scACoCcwGbjdXR4JJIX679hejb+iMCY4\n3lPVKhFZinMyeN9dvhTnZNIPOBD40L1oj8TpSbNGTQdgC93twTmZ3y4i2cDrqvpjnWOOBqapqgen\n47BPgMOAIuAbVc0DEJHv3H1+XvNBEUkBUlX1E3fRC/zy3l+vEZHT3ekcoA+QX2ebY4BDgfnu945n\nT6dmlTgn7JrvfVwAx3wZJxnMwenG/J/u8gNF5K9AKtABmOX3mVfd36iubOBlt8QWA/zkt26GqpYB\nZW6JaTjwnd/644GDROQsdz4F5/vPB551S4dvqqr/Z0wYsuojs8/E6YfJQ/09NVYAqNNDZpW6l4o4\n/bhEAQIsU6eNYoiqDlbV4+t+3t1/lLuvl3CqUsqAd0XkN78g3Aq/ad8+m4uIHAUcC4xU1YOBb4G4\n+jYFXvD73v1UdYq7zv93CjTGmcAYEUnHSTaz3eXPA5NUdTBwZ51YShrY12M4pYbBwOV1PlO3P5y6\n84JTGqn5Xj1V9QN1Bt/5Nc7ofM+LyIUBfCcTQpYUzD4RkSzgSZyTyL50oLUKyBKRke7+okVkUBPH\n7AWsVdVHcapNDqqzyWfAOW79dhbOyeibQIJRp1O/XX513hcAnzTykbpSgF2qWioi/YHD/dZV1bSj\nAB8DZ4lIR/c7pYtI9yb2vRtIaiDuYpyr8Udwqu9qSgBJwBb3uL/9Bd9hkzt9UZ11p4lInIhk4FTH\nza+zfhZOu0tNe1Fft42jO7BVVf8FPA0cEmAsJkSs+sj8EvFu1Us0UA38B9inrshVtdKtanjUrbqJ\nAh4GljXysbOBC0SkCmfUqXvqrH8DGIlT/63AH1T1Z/ckHYiLgCfdRtm1wMUBfyGneuz3IrICJ+F9\n5bfuKWCJiCxS1d+KyB9xRs6LAKqAq3B6vGzIdOBfInINTttC3TugXsap6jrKb9mfcHrU3O6+15tU\n6pgCvCoiu3BKHD391i3BqaLKBP6iqpvFGRGsxtM4VXKL3Eb87ThtR0cBN7n/ZsWAlRTCnPWSaowx\nxseqj4wxxvhYUjDGGONjScEYY4yPJQVjjDE+lhSMMcb4WFIwxhjjY0nBGGOMjyUFY4wxPv8fSa9X\n83nvdgkAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f95bd6914d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "N = 8\n",
    "\n",
    "ind = np.arange(N)  # the x locations for the groups\n",
    "width = 0.35       # the width of the bars\n",
    "\n",
    "label = [\"ASW\", \"ARI\", \"NMI\"]\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "for i in [0, 1, 2]:\n",
    "    ax.errorbar(latent_d, np.mean(res[:, :, i], axis=0), yerr=np.std(res[:, :, i], axis=0), label=label[i])\n",
    "\n",
    "# add some text for labels, title and axes ticks\n",
    "ax.set_ylabel('Clustering metrics')\n",
    "ax.set_xticks(latent_d)\n",
    "ax.set_xlabel(\"Dimension of latent variables\")\n",
    "plt.legend()\n",
    "\n",
    "plt.savefig(\"zeiler_clustering_dimension_stability.pdf\", dpi=300)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}