BostonSingular.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-20T20:09:48.283910Z",
     "start_time": "2019-01-20T20:09:46.461938Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Initialized TensorFlow\n"
     ]
    }
   ],
   "source": [
    "#!rm -rf code\n",
    "#!git clone https://xxxx/yyyy.git code\n",
    "#!mv code/* .\n",
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline\n",
    "from helpers import *\n",
    "from matplotlib import pyplot as plt\n",
    "import numpy as np\n",
    "from experiment_datasets import *\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import pickle"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Fixed single Boston experiment for all dataset as points"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-20T20:09:48.298982Z",
     "start_time": "2019-01-20T20:09:48.285384Z"
    }
   },
   "outputs": [],
   "source": [
    "# pfailure for the first layer\n",
    "p = 1e-2\n",
    "np.random.seed(42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-20T20:09:50.215678Z",
     "start_time": "2019-01-20T20:09:48.300332Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_10 (Dense)             (None, 50)                700       \n",
      "_________________________________________________________________\n",
      "dense_11 (Dense)             (None, 49)                2499      \n",
      "_________________________________________________________________\n",
      "dense_12 (Dense)             (None, 1)                 50        \n",
      "=================================================================\n",
      "Total params: 3,249\n",
      "Trainable params: 3,249\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Train on 404 samples, validate on 102 samples\n",
      "Epoch 1/50\n",
      "404/404 [==============================] - 0s 424us/step - loss: 531.2130 - categorical_accuracy: 1.0000 - mean_squared_error: 531.2130 - mean_absolute_error: 21.1292 - val_loss: 554.3857 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 554.3857 - val_mean_absolute_error: 21.6956\n",
      "Epoch 2/50\n",
      "404/404 [==============================] - 0s 7us/step - loss: 525.8105 - categorical_accuracy: 1.0000 - mean_squared_error: 525.8105 - mean_absolute_error: 20.9983 - val_loss: 548.7767 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 548.7767 - val_mean_absolute_error: 21.5615\n",
      "Epoch 3/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 520.2533 - categorical_accuracy: 1.0000 - mean_squared_error: 520.2533 - mean_absolute_error: 20.8623 - val_loss: 542.9644 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 542.9644 - val_mean_absolute_error: 21.4190\n",
      "Epoch 4/50\n",
      "404/404 [==============================] - 0s 6us/step - loss: 514.6258 - categorical_accuracy: 1.0000 - mean_squared_error: 514.6258 - mean_absolute_error: 20.7233 - val_loss: 536.6470 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 536.6470 - val_mean_absolute_error: 21.2657\n",
      "Epoch 5/50\n",
      "404/404 [==============================] - 0s 6us/step - loss: 508.6155 - categorical_accuracy: 1.0000 - mean_squared_error: 508.6155 - mean_absolute_error: 20.5735 - val_loss: 530.2158 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 530.2158 - val_mean_absolute_error: 21.1106\n",
      "Epoch 6/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 502.3077 - categorical_accuracy: 1.0000 - mean_squared_error: 502.3077 - mean_absolute_error: 20.4181 - val_loss: 523.3417 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 523.3417 - val_mean_absolute_error: 20.9435\n",
      "Epoch 7/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 495.6501 - categorical_accuracy: 1.0000 - mean_squared_error: 495.6501 - mean_absolute_error: 20.2517 - val_loss: 516.4916 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 516.4916 - val_mean_absolute_error: 20.7776\n",
      "Epoch 8/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 488.7649 - categorical_accuracy: 1.0000 - mean_squared_error: 488.7649 - mean_absolute_error: 20.0825 - val_loss: 508.8617 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 508.8617 - val_mean_absolute_error: 20.5894\n",
      "Epoch 9/50\n",
      "404/404 [==============================] - 0s 6us/step - loss: 481.5795 - categorical_accuracy: 1.0000 - mean_squared_error: 481.5795 - mean_absolute_error: 19.9002 - val_loss: 500.7063 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 500.7063 - val_mean_absolute_error: 20.3876\n",
      "Epoch 10/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 473.4600 - categorical_accuracy: 1.0000 - mean_squared_error: 473.4600 - mean_absolute_error: 19.6958 - val_loss: 491.0345 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 491.0345 - val_mean_absolute_error: 20.1445\n",
      "Epoch 11/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 464.1482 - categorical_accuracy: 1.0000 - mean_squared_error: 464.1482 - mean_absolute_error: 19.4533 - val_loss: 478.5085 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 478.5085 - val_mean_absolute_error: 19.8242\n",
      "Epoch 12/50\n",
      "404/404 [==============================] - 0s 6us/step - loss: 452.5789 - categorical_accuracy: 1.0000 - mean_squared_error: 452.5789 - mean_absolute_error: 19.1451 - val_loss: 462.1184 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 462.1184 - val_mean_absolute_error: 19.3995\n",
      "Epoch 13/50\n",
      "404/404 [==============================] - 0s 6us/step - loss: 437.0303 - categorical_accuracy: 1.0000 - mean_squared_error: 437.0303 - mean_absolute_error: 18.7274 - val_loss: 441.9381 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 441.9381 - val_mean_absolute_error: 18.8642\n",
      "Epoch 14/50\n",
      "404/404 [==============================] - 0s 6us/step - loss: 417.5897 - categorical_accuracy: 1.0000 - mean_squared_error: 417.5897 - mean_absolute_error: 18.1932 - val_loss: 422.8847 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 422.8847 - val_mean_absolute_error: 18.3610\n",
      "Epoch 15/50\n",
      "404/404 [==============================] - 0s 6us/step - loss: 399.1344 - categorical_accuracy: 1.0000 - mean_squared_error: 399.1344 - mean_absolute_error: 17.6874 - val_loss: 406.6974 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 406.6974 - val_mean_absolute_error: 17.9266\n",
      "Epoch 16/50\n",
      "404/404 [==============================] - 0s 6us/step - loss: 383.3653 - categorical_accuracy: 1.0000 - mean_squared_error: 383.3653 - mean_absolute_error: 17.2566 - val_loss: 391.7862 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 391.7862 - val_mean_absolute_error: 17.5224\n",
      "Epoch 17/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 369.1382 - categorical_accuracy: 1.0000 - mean_squared_error: 369.1382 - mean_absolute_error: 16.8532 - val_loss: 380.0009 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 380.0009 - val_mean_absolute_error: 17.1930\n",
      "Epoch 18/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 357.7100 - categorical_accuracy: 1.0000 - mean_squared_error: 357.7100 - mean_absolute_error: 16.5176 - val_loss: 369.6230 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 369.6230 - val_mean_absolute_error: 16.8931\n",
      "Epoch 19/50\n",
      "404/404 [==============================] - 0s 6us/step - loss: 347.1129 - categorical_accuracy: 1.0000 - mean_squared_error: 347.1129 - mean_absolute_error: 16.2019 - val_loss: 359.7038 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 359.7038 - val_mean_absolute_error: 16.6084\n",
      "Epoch 20/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 337.8010 - categorical_accuracy: 1.0000 - mean_squared_error: 337.8010 - mean_absolute_error: 15.9161 - val_loss: 350.6469 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 350.6469 - val_mean_absolute_error: 16.3407\n",
      "Epoch 21/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 329.1826 - categorical_accuracy: 1.0000 - mean_squared_error: 329.1826 - mean_absolute_error: 15.6472 - val_loss: 341.7916 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 341.7916 - val_mean_absolute_error: 16.0812\n",
      "Epoch 22/50\n",
      "404/404 [==============================] - 0s 6us/step - loss: 321.0450 - categorical_accuracy: 1.0000 - mean_squared_error: 321.0450 - mean_absolute_error: 15.3909 - val_loss: 333.9474 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 333.9474 - val_mean_absolute_error: 15.8511\n",
      "Epoch 23/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 313.5004 - categorical_accuracy: 1.0000 - mean_squared_error: 313.5004 - mean_absolute_error: 15.1513 - val_loss: 326.4434 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 326.4434 - val_mean_absolute_error: 15.6341\n",
      "Epoch 24/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 306.4294 - categorical_accuracy: 1.0000 - mean_squared_error: 306.4294 - mean_absolute_error: 14.9233 - val_loss: 319.4004 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 319.4004 - val_mean_absolute_error: 15.4295\n",
      "Epoch 25/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 299.6923 - categorical_accuracy: 1.0000 - mean_squared_error: 299.6923 - mean_absolute_error: 14.7025 - val_loss: 312.6623 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 312.6623 - val_mean_absolute_error: 15.2308\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 26/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 293.2552 - categorical_accuracy: 1.0000 - mean_squared_error: 293.2552 - mean_absolute_error: 14.4888 - val_loss: 306.2101 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 306.2101 - val_mean_absolute_error: 15.0379\n",
      "Epoch 27/50\n",
      "404/404 [==============================] - 0s 6us/step - loss: 287.0847 - categorical_accuracy: 1.0000 - mean_squared_error: 287.0847 - mean_absolute_error: 14.2812 - val_loss: 299.9983 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 299.9983 - val_mean_absolute_error: 14.8523\n",
      "Epoch 28/50\n",
      "404/404 [==============================] - 0s 6us/step - loss: 281.1357 - categorical_accuracy: 1.0000 - mean_squared_error: 281.1357 - mean_absolute_error: 14.0786 - val_loss: 293.9557 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 293.9557 - val_mean_absolute_error: 14.6713\n",
      "Epoch 29/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 275.3815 - categorical_accuracy: 1.0000 - mean_squared_error: 275.3815 - mean_absolute_error: 13.8817 - val_loss: 288.2031 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 288.2031 - val_mean_absolute_error: 14.4959\n",
      "Epoch 30/50\n",
      "404/404 [==============================] - 0s 6us/step - loss: 269.8818 - categorical_accuracy: 1.0000 - mean_squared_error: 269.8818 - mean_absolute_error: 13.6932 - val_loss: 282.6442 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 282.6442 - val_mean_absolute_error: 14.3239\n",
      "Epoch 31/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 264.5815 - categorical_accuracy: 1.0000 - mean_squared_error: 264.5815 - mean_absolute_error: 13.5101 - val_loss: 277.2662 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 277.2662 - val_mean_absolute_error: 14.1552\n",
      "Epoch 32/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 259.4770 - categorical_accuracy: 1.0000 - mean_squared_error: 259.4770 - mean_absolute_error: 13.3327 - val_loss: 272.0575 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 272.0575 - val_mean_absolute_error: 13.9896\n",
      "Epoch 33/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 254.5381 - categorical_accuracy: 1.0000 - mean_squared_error: 254.5381 - mean_absolute_error: 13.1590 - val_loss: 267.0087 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 267.0087 - val_mean_absolute_error: 13.8268\n",
      "Epoch 34/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 249.7548 - categorical_accuracy: 1.0000 - mean_squared_error: 249.7548 - mean_absolute_error: 12.9883 - val_loss: 262.1115 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 262.1115 - val_mean_absolute_error: 13.6695\n",
      "Epoch 35/50\n",
      "404/404 [==============================] - 0s 6us/step - loss: 245.1189 - categorical_accuracy: 1.0000 - mean_squared_error: 245.1189 - mean_absolute_error: 12.8205 - val_loss: 257.3588 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 257.3588 - val_mean_absolute_error: 13.5191\n",
      "Epoch 36/50\n",
      "404/404 [==============================] - 0s 6us/step - loss: 240.6232 - categorical_accuracy: 1.0000 - mean_squared_error: 240.6232 - mean_absolute_error: 12.6562 - val_loss: 252.7442 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 252.7442 - val_mean_absolute_error: 13.3710\n",
      "Epoch 37/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 236.2616 - categorical_accuracy: 1.0000 - mean_squared_error: 236.2616 - mean_absolute_error: 12.4946 - val_loss: 248.2620 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 248.2620 - val_mean_absolute_error: 13.2268\n",
      "Epoch 38/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 232.0285 - categorical_accuracy: 1.0000 - mean_squared_error: 232.0285 - mean_absolute_error: 12.3360 - val_loss: 243.9069 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 243.9069 - val_mean_absolute_error: 13.0880\n",
      "Epoch 39/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 227.9187 - categorical_accuracy: 1.0000 - mean_squared_error: 227.9187 - mean_absolute_error: 12.1804 - val_loss: 239.6744 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 239.6744 - val_mean_absolute_error: 12.9532\n",
      "Epoch 40/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 223.9276 - categorical_accuracy: 1.0000 - mean_squared_error: 223.9276 - mean_absolute_error: 12.0291 - val_loss: 235.5600 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 235.5600 - val_mean_absolute_error: 12.8204\n",
      "Epoch 41/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 220.0509 - categorical_accuracy: 1.0000 - mean_squared_error: 220.0509 - mean_absolute_error: 11.8808 - val_loss: 231.5596 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 231.5596 - val_mean_absolute_error: 12.6896\n",
      "Epoch 42/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 216.2847 - categorical_accuracy: 1.0000 - mean_squared_error: 216.2847 - mean_absolute_error: 11.7353 - val_loss: 227.6696 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 227.6696 - val_mean_absolute_error: 12.5607\n",
      "Epoch 43/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 212.6252 - categorical_accuracy: 1.0000 - mean_squared_error: 212.6252 - mean_absolute_error: 11.5931 - val_loss: 223.8862 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 223.8862 - val_mean_absolute_error: 12.4336\n",
      "Epoch 44/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 209.0690 - categorical_accuracy: 1.0000 - mean_squared_error: 209.0690 - mean_absolute_error: 11.4544 - val_loss: 220.2062 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 220.2062 - val_mean_absolute_error: 12.3083\n",
      "Epoch 45/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 205.6128 - categorical_accuracy: 1.0000 - mean_squared_error: 205.6128 - mean_absolute_error: 11.3186 - val_loss: 216.6263 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 216.6263 - val_mean_absolute_error: 12.1849\n",
      "Epoch 46/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 202.2535 - categorical_accuracy: 1.0000 - mean_squared_error: 202.2535 - mean_absolute_error: 11.1853 - val_loss: 213.1434 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 213.1434 - val_mean_absolute_error: 12.0631\n",
      "Epoch 47/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 198.9881 - categorical_accuracy: 1.0000 - mean_squared_error: 198.9881 - mean_absolute_error: 11.0542 - val_loss: 209.7547 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 209.7547 - val_mean_absolute_error: 11.9431\n",
      "Epoch 48/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 195.8137 - categorical_accuracy: 1.0000 - mean_squared_error: 195.8137 - mean_absolute_error: 10.9255 - val_loss: 206.4574 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 206.4574 - val_mean_absolute_error: 11.8249\n",
      "Epoch 49/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 192.7277 - categorical_accuracy: 1.0000 - mean_squared_error: 192.7277 - mean_absolute_error: 10.8013 - val_loss: 203.2487 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 203.2487 - val_mean_absolute_error: 11.7110\n",
      "Epoch 50/50\n",
      "404/404 [==============================] - 0s 5us/step - loss: 189.7274 - categorical_accuracy: 1.0000 - mean_squared_error: 189.7274 - mean_absolute_error: 10.6802 - val_loss: 200.1262 - val_categorical_accuracy: 1.0000 - val_mean_squared_error: 200.1262 - val_mean_absolute_error: 11.5988\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Creating network for 13-dimensional input and 1-dimensional output, with 2 hidden layers\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_13 (Dense)             (None, 50)                700       \n",
      "_________________________________________________________________\n",
      "Crashes (Lambda)             (None, 50)                0         \n",
      "_________________________________________________________________\n",
      "dense_14 (Dense)             (None, 49)                2499      \n",
      "_________________________________________________________________\n",
      "dense_15 (Dense)             (None, 1)                 50        \n",
      "=================================================================\n",
      "Total params: 3,249\n",
      "Trainable params: 3,249\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_16 (Dense)             (None, 50)                700       \n",
      "_________________________________________________________________\n",
      "dense_17 (Dense)             (None, 49)                2499      \n",
      "_________________________________________________________________\n",
      "dense_18 (Dense)             (None, 1)                 50        \n",
      "=================================================================\n",
      "Total params: 3,249\n",
      "Trainable params: 3,249\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# run with random weights and multiple inputs\n",
    "exp = BostonHousingExperiment(N = [50, 49], p_inference = [0, p, 0, 0], seed = 2, activation = 'sigmoid', epochs = 50, do_print = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAADbhJREFUeJzt3XGsnfVdx/H3hxacEwbG3hmk7Upi0TWoA28YCSZiYFlhSfvH1LRmTglZExWDgRjRKSr7RyTBxKzKGrcwFx3r5qJX6axmgiTGYi+Owdra5aZDKFtCYQydhGHd1z/O2XJ2uO15bvvce+iv71dyk/M855dzvr+UvHv6nHsOqSokSW05Z9oDSJL6Z9wlqUHGXZIaZNwlqUHGXZIaZNwlqUHGXZIaZNwlqUHGXZIatHpaT7xmzZrasGHDtJ5eks5Ijz322PNVNTNp3dTivmHDBubn56f19JJ0Rkryn13WeVlGkhpk3CWpQcZdkhpk3CWpQcZdkho0Me5JPpLkuSRfOMH9SfLHSRaSPJHkyv7HlCQtRZdX7vcDm09y/w3AxuHPDuBPT38sSdLpmBj3qnoE+OpJlmwF/rwG9gEXJbm4rwElSUvXxzX3S4BnRo6PDs9JkqZkRT+hmmQHg0s3rF+/fiWfWupswx0PTu25n/qDd03tudWWPl65PwusGzleOzz3GlW1q6pmq2p2ZmbiVyNIkk5RH3GfA947/K2Zq4GXquorPTyuJOkUTbwsk+TjwLXAmiRHgd8FzgWoqvuAPcCNwALwMnDTcg0rSepmYtyravuE+wv4ld4mkiSdNj+hKkkNMu6S1CDjLkkNMu6S1CDjLkkNMu6S1CDjLkkNMu6S1CDjLkkNMu6S1CDjLkkNMu6S1CDjLkkNMu6S1CDjLkkNMu6S1CDjLkkNMu6S1CDjLkkNMu6S1CDjLkkNMu6S1CDjLkkNMu6S1CDjLkkNMu6S1CDjLkkNMu6S1CDjLkkNMu6S1CDjLkkNMu6S1KBOcU+yOcnhJAtJ7ljk/vVJHkryuSRPJLmx/1ElSV1NjHuSVcBO4AZgE7A9yaaxZb8N7K6qK4BtwJ/0Pagkqbsur9yvAhaq6khVvQo8AGwdW1PAm4a3LwS+3N+IkqSlWt1hzSXAMyPHR4G3j635PeAfkvwq8D3A9b1MJ0k6JX29oboduL+q1gI3Ah9L8prHTrIjyXyS+WPHjvX01JKkcV3i/iywbuR47fDcqJuB3QBV9a/AG4A14w9UVbuqaraqZmdmZk5tYknSRF3ivh/YmOTSJOcxeMN0bmzN08B1AEneyiDuvjSXpCmZGPeqOg7cAuwFDjH4rZgDSe5KsmW47HbgfUk+D3wc+MWqquUaWpJ0cl3eUKWq9gB7xs7dOXL7IHBNv6NJkk6Vn1CVpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAZ1inuSzUkOJ1lIcscJ1vxskoNJDiT5y37HlCQtxepJC5KsAnYC7wCOAvuTzFXVwZE1G4HfBK6pqheTvHm5BpYkTdbllftVwEJVHamqV4EHgK1ja94H7KyqFwGq6rl+x5QkLUWXuF8CPDNyfHR4btRlwGVJ/iXJviSb+xpQkrR0Ey/LLOFxNgLXAmuBR5L8SFV9bXRRkh3ADoD169f39NSSpHFdXrk/C6wbOV47PDfqKDBXVf9bVV8Cvsgg9t+hqnZV1WxVzc7MzJzqzJKkCbrEfT+wMcmlSc4DtgFzY2v+msGrdpKsYXCZ5kiPc0qSlmBi3KvqOHALsBc4BOyuqgNJ7kqyZbhsL/BCkoPAQ8CvV9ULyzW0JOnkOl1zr6o9wJ6xc3eO3C7gtuGPJGnK/ISqJDXIuEtSg4y7JDXIuEtSg4y7JDXIuEtSg4y7JDXIuEtSg4y7JDXIuEtSg4y7JDXIuEtSg4y7JDXIuEtSg4y7JDXIuEtSg4y7JDXIuEtSg4y7JDXIuEtSg4y7JDXIuEtSg4y7JDXIuEtSg4y7JDXIuEtSg4y7JDXIuEtSg4y7JDXIuEtSg4y7JDXIuEtSgzrFPcnmJIeTLCS54yTr3p2kksz2N6Ikaakmxj3JKmAncAOwCdieZNMi6y4AbgUe7XtISdLSdHnlfhWwUFVHqupV4AFg6yLrPgDcDbzS43ySpFPQJe6XAM+MHB8dnvu2JFcC66rqwR5nkySdotN+QzXJOcC9wO0d1u5IMp9k/tixY6f71JKkE+gS92eBdSPHa4fnvuUC4HLg4SRPAVcDc4u9qVpVu6pqtqpmZ2ZmTn1qSdJJdYn7fmBjkkuTnAdsA+a+dWdVvVRVa6pqQ1VtAPYBW6pqflkmliRNNDHuVXUcuAXYCxwCdlfVgSR3Jdmy3ANKkpZudZdFVbUH2DN27s4TrL329MeSJJ0OP6EqSQ0y7pLUIOMuSQ0y7pLUIOMuSQ0y7pLUIOMuSQ0y7pLUIOMuSQ0y7pLUIOMuSQ0y7pLUIOMuSQ0y7pLUIOMuSQ0y7pLUIOMuSQ0y7pLUIOMuSQ0y7pLUIOMuSQ0y7pLUIOMuSQ0y7pLUIOMuSQ0y7pLUIOMuSQ0y7pLUIOMuSQ0y7pLUIOMuSQ0y7pLUIOMuSQ3qFPckm5McTrKQ5I5F7r8tycEkTyT5bJK39D+qJKmriXFPsgrYCdwAbAK2J9k0tuxzwGxV/SjwKeAP+x5UktRdl1fuVwELVXWkql4FHgC2ji6oqoeq6uXh4T5gbb9jSpKWokvcLwGeGTk+Ojx3IjcDn1nsjiQ7kswnmT927Fj3KSVJS9LrG6pJ3gPMAvcsdn9V7aqq2aqanZmZ6fOpJUkjVndY8yywbuR47fDcd0hyPfB+4Cer6hv9jCdJOhVdXrnvBzYmuTTJecA2YG50QZIrgA8BW6rquf7HlCQtxcS4V9Vx4BZgL3AI2F1VB5LclWTLcNk9wPnAJ5M8nmTuBA8nSVoBXS7LUFV7gD1j5+4cuX19z3NJkk6Dn1CVpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqkHGXpAYZd0lqUKe4J9mc5HCShSR3LHL/dyX5xPD+R5Ns6HtQSVJ3E+OeZBWwE7gB2ARsT7JpbNnNwItV9YPAHwF39z2oJKm7Lq/crwIWqupIVb0KPABsHVuzFfjo8PangOuSpL8xJUlL0SXulwDPjBwfHZ5bdE1VHQdeAr6vjwElSUu3eiWfLMkOYMfw8OtJDq/k8/dsDfD8tIdYQWfbfmEKe870L2j65/z695Yui7rE/Vlg3cjx2uG5xdYcTbIauBB4YfyBqmoXsKvLYK93Searanbac6yUs22/4J7PFq3uuctlmf3AxiSXJjkP2AbMja2ZA35hePungX+qqupvTEnSUkx85V5Vx5PcAuwFVgEfqaoDSe4C5qtqDvgw8LEkC8BXGfwFIEmakk7X3KtqD7Bn7NydI7dfAX6m39Fe95q4vLQEZ9t+wT2fLZrcc7x6Iknt8esHJKlBxv0EkqxL8lCSg0kOJLn1BOuuTfL4cM0/r/Scfeqy5yQXJvnbJJ8frrlpGrP2JckbkvzbyH5+f5E1TX29Rsc93zb87+CJJJ9N0unX716vuux5ZO27k1SSM/s3aKrKn0V+gIuBK4e3LwC+CGwaW3MRcBBYPzx+87TnXoE9/xZw9/D2DIM30M+b9uynsecA5w9vnws8Clw9tuaXgfuGt7cBn5j23Cuw558C3ji8/Utnw56H910APALsA2anPffp/PjK/QSq6itV9e/D2/8NHOK1n8z9OeDTVfX0cN1zKztlvzruuYALhl8vcT6DuB9f0UF7VANfHx6eO/wZfyOqqa/X6LLnqnqoql4eHu5j8PmWM1bHP2eADzD4bqxXVmq25WLcOxj+M/wKBn/bj7oM+N4kDyd5LMl7V3q25XKSPX8QeCvwZeBJ4Naq+uaKDtezJKuSPA48B/xjVY3vubmv1+iw51E3A59ZmcmWz6Q9J7kSWFdVD05lwJ4Z9wmSnA/8FfBrVfVfY3evBn4ceBfwTuB3kly2wiP2bsKe3wk8DvwA8Dbgg0netMIj9qqq/q+q3sbg1elVSS6f9kzLreuek7wHmAXuWcn5lsPJ9pzkHOBe4PZpzdc3434SSc5lELm/qKpPL7LkKLC3qv6nqp5ncK3ux1Zyxr512PNNDC5FVVUtAF8CfnglZ1wuVfU14CFg89hd3/4KjpN9vcaZ6CR7Jsn1wPuBLVX1jZWebbmcYM8XAJcDDyd5CrgamDuT31Q17icwvKb6YeBQVd17gmV/A/xEktVJ3gi8ncF16jNSxz0/DVw3XP/9wA8BR1Zmwv4lmUly0fD2dwPvAP5jbFlTX6/RZc9JrgA+xCDsZ/R7STB5z1X1UlWtqaoNVbWBwfsMW6pqfioD92BFvxXyDHMN8PPAk8PrdDD4TZH1AFV1X1UdSvL3wBPAN4E/q6ovTGXafkzcM4M3nO5P8iSD30D4jeG/Ws5UFwMfHf5Pac4BdlfV3zX+9Rpd9nwPgzfMPzl87/jpqtoytYlPX5c9N8VPqEpSg7wsI0kNMu6S1CDjLkkNMu6S1CDjLkkNMu6S1CDjLkkNMu6S1KD/B5q9gfCwjVTOAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "for i in range(len(exp.W)):\n",
    "    plt.hist(np.linalg.svd(exp.W[i])[1])\n",
    "    plt.show()"
   ]
  }
 ],
 "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": "3.6.8"
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}