variational_autoencoder_ukbb.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from scipy.stats import norm\n",
    "\n",
    "from keras.layers import Input, Dense, Lambda, Layer, Dropout\n",
    "from keras.models import Model\n",
    "from keras import backend as K\n",
    "from keras import metrics\n",
    "import pandas as pd\n",
    "#import tensorflow as tf\n",
    "import keras"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import matplotlib.pyplot as plt\n",
    "plt.style.use('ggplot')\n",
    "%matplotlib inline\n",
    "#import pydot\n",
    "#import graphviz\n",
    "from keras.utils import plot_model\n",
    "from IPython.display import SVG\n",
    "from keras.utils.vis_utils import model_to_dot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn import preprocessing\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn import manifold"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/wakari/anaconda/envs/default/lib/python2.7/site-packages/IPython/core/interactiveshell.py:2714: DtypeWarning: Columns (3,2931,2943) have mixed types. Specify dtype option on import or set low_memory=False.\n",
      "  interactivity=interactivity, compiler=compiler, result=result)\n"
     ]
    }
   ],
   "source": [
    "asthma_df=pd.read_csv(\"ukbb_asthma_sample_5k.csv\",sep='\\t')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Select columns to filter on\n",
    "cols = [col for col in asthma_df.columns if 'HES_p' not in col and 'PC' not in col]\n",
    "asthma_df=asthma_df[cols].drop('Unnamed: 0',axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#select only 'QUANTITY' fields\n",
    "asthma_df_quant=asthma_df.loc[:, asthma_df.columns.str.contains('QUANT|age|BMI')].fillna(0.0)\n",
    "#select only fields with 10 or more unique values\n",
    "for col in asthma_df_quant.columns:\n",
    "    if len(asthma_df_quant[col].unique()) < 20 :\n",
    "        asthma_df_quant.drop(col,inplace=True,axis=1)\n",
    "\n",
    "        \n",
    "#select only 'QUANTITY' and \"CATEGORY\" fields\n",
    "asthma_df_quant_cat=asthma_df.loc[:, asthma_df.columns.str.contains('QUANT|age|BMI|CAT|sex')].fillna(0.0)\n",
    "asthma_df_quant_cat=asthma_df_quant_cat.drop(columns=['f_22182_0_0_f_CAT_HLA_imputation_values_and_quality'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#asthma_df_quant_cat['f_40019_0_p_D44_CAT_Neoplasm_of_uncertain_or_unknown_behaviour_of_endocrine_glands'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#asthma_df_quant_cat.loc[:,asthma_df_quant_cat.columns.str.contains('CAT')]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "      <th>BMI</th>\n",
       "      <th>f_20006_0_0_f_QUANT_Interpolated_Year_when_cancer_first_diagnosed</th>\n",
       "      <th>f_20008_0_0_f_QUANT_Interpolated_Year_when_non_cancer_illness_first_diagnosed</th>\n",
       "      <th>f_20010_0_0_f_QUANT_Interpolated_Year_when_operation_took_place</th>\n",
       "      <th>f_1269_0_0_f_QUANT_Exposure_to_tobacco_smoke_at_home</th>\n",
       "      <th>f_1279_0_0_f_QUANT_Exposure_to_tobacco_smoke_outside_home</th>\n",
       "      <th>f_1807_0_0_f_QUANT_Father_s_age_at_death</th>\n",
       "      <th>f_2867_0_0_f_QUANT_Age_started_smoking_in_former_smokers</th>\n",
       "      <th>f_2966_0_0_f_QUANT_Age_high_blood_pressure_diagnosed</th>\n",
       "      <th>...</th>\n",
       "      <th>specialRequest_QUANT_number_of_hospitalizing_copd_exacerbation_caseOnly_includes_SPIROMETRY_NI_HES_residualizedRelated</th>\n",
       "      <th>specialRequest_QUANT_number_of_hospitalizing_copd_exacerbation_caseOnly_smokersOnly_includes_SPIROMETRY_NI_HES_residualizedRelated</th>\n",
       "      <th>specialRequest_QUANT_numCigarettes_residualizedRelated</th>\n",
       "      <th>specialRequest_QUANT_age_started_smoking_residualizedRelated</th>\n",
       "      <th>specialRequest_QUANT_age_stopped_smoking_residualizedRelated</th>\n",
       "      <th>specialRequest_QUANT_pack_years_residualizedRelated</th>\n",
       "      <th>specialRequest_QUANT_appendicular_lean_mass_residualizedRelated</th>\n",
       "      <th>specialRequest_ageSquared_QUANT_age_squared</th>\n",
       "      <th>specialRequest_QUANT_years_observed_in_primary_HES</th>\n",
       "      <th>specialRequest_QUANT_OA_HES_based_age_of_onset</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>63</td>\n",
       "      <td>21.3476</td>\n",
       "      <td>2004.5</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1980.5</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>60.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>3969</td>\n",
       "      <td>0.134247</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>51</td>\n",
       "      <td>25.9862</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2007.50</td>\n",
       "      <td>1965.5</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>2601</td>\n",
       "      <td>0.002740</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>58</td>\n",
       "      <td>21.2239</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2010.23</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>75.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>58.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>3364</td>\n",
       "      <td>0.002740</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>69</td>\n",
       "      <td>27.0389</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1970.88</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>59.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>30.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>4761</td>\n",
       "      <td>0.002740</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>64</td>\n",
       "      <td>31.5993</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1996.17</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>89.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.132985</td>\n",
       "      <td>4096</td>\n",
       "      <td>0.010959</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 638 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   age      BMI  \\\n",
       "0   63  21.3476   \n",
       "1   51  25.9862   \n",
       "2   58  21.2239   \n",
       "3   69  27.0389   \n",
       "4   64  31.5993   \n",
       "\n",
       "   f_20006_0_0_f_QUANT_Interpolated_Year_when_cancer_first_diagnosed  \\\n",
       "0                                             2004.5                   \n",
       "1                                                0.0                   \n",
       "2                                                0.0                   \n",
       "3                                                0.0                   \n",
       "4                                                0.0                   \n",
       "\n",
       "   f_20008_0_0_f_QUANT_Interpolated_Year_when_non_cancer_illness_first_diagnosed  \\\n",
       "0                                               0.00                               \n",
       "1                                            2007.50                               \n",
       "2                                            2010.23                               \n",
       "3                                            1970.88                               \n",
       "4                                            1996.17                               \n",
       "\n",
       "   f_20010_0_0_f_QUANT_Interpolated_Year_when_operation_took_place  \\\n",
       "0                                             1980.5                 \n",
       "1                                             1965.5                 \n",
       "2                                                0.0                 \n",
       "3                                                0.0                 \n",
       "4                                                0.0                 \n",
       "\n",
       "   f_1269_0_0_f_QUANT_Exposure_to_tobacco_smoke_at_home  \\\n",
       "0                                                0.0      \n",
       "1                                                0.0      \n",
       "2                                                0.0      \n",
       "3                                                0.0      \n",
       "4                                                0.0      \n",
       "\n",
       "   f_1279_0_0_f_QUANT_Exposure_to_tobacco_smoke_outside_home  \\\n",
       "0                                                0.0           \n",
       "1                                                1.0           \n",
       "2                                                0.0           \n",
       "3                                                0.0           \n",
       "4                                                0.0           \n",
       "\n",
       "   f_1807_0_0_f_QUANT_Father_s_age_at_death  \\\n",
       "0                                      60.0   \n",
       "1                                       0.0   \n",
       "2                                      75.0   \n",
       "3                                      59.0   \n",
       "4                                      89.0   \n",
       "\n",
       "   f_2867_0_0_f_QUANT_Age_started_smoking_in_former_smokers  \\\n",
       "0                                                0.0          \n",
       "1                                                0.0          \n",
       "2                                                0.0          \n",
       "3                                                0.0          \n",
       "4                                                0.0          \n",
       "\n",
       "   f_2966_0_0_f_QUANT_Age_high_blood_pressure_diagnosed  \\\n",
       "0                                                0.0      \n",
       "1                                                0.0      \n",
       "2                                               58.0      \n",
       "3                                               30.0      \n",
       "4                                                0.0      \n",
       "\n",
       "                        ...                        \\\n",
       "0                       ...                         \n",
       "1                       ...                         \n",
       "2                       ...                         \n",
       "3                       ...                         \n",
       "4                       ...                         \n",
       "\n",
       "   specialRequest_QUANT_number_of_hospitalizing_copd_exacerbation_caseOnly_includes_SPIROMETRY_NI_HES_residualizedRelated  \\\n",
       "0                                                0.0                                                                        \n",
       "1                                                0.0                                                                        \n",
       "2                                                0.0                                                                        \n",
       "3                                                0.0                                                                        \n",
       "4                                                0.0                                                                        \n",
       "\n",
       "   specialRequest_QUANT_number_of_hospitalizing_copd_exacerbation_caseOnly_smokersOnly_includes_SPIROMETRY_NI_HES_residualizedRelated  \\\n",
       "0                                                0.0                                                                                    \n",
       "1                                                0.0                                                                                    \n",
       "2                                                0.0                                                                                    \n",
       "3                                                0.0                                                                                    \n",
       "4                                                0.0                                                                                    \n",
       "\n",
       "   specialRequest_QUANT_numCigarettes_residualizedRelated  \\\n",
       "0                                                0.0        \n",
       "1                                                0.0        \n",
       "2                                                0.0        \n",
       "3                                                0.0        \n",
       "4                                                0.0        \n",
       "\n",
       "   specialRequest_QUANT_age_started_smoking_residualizedRelated  \\\n",
       "0                                                0.0              \n",
       "1                                                0.0              \n",
       "2                                                0.0              \n",
       "3                                                0.0              \n",
       "4                                                0.0              \n",
       "\n",
       "   specialRequest_QUANT_age_stopped_smoking_residualizedRelated  \\\n",
       "0                                                0.0              \n",
       "1                                                0.0              \n",
       "2                                                0.0              \n",
       "3                                                0.0              \n",
       "4                                                0.0              \n",
       "\n",
       "   specialRequest_QUANT_pack_years_residualizedRelated  \\\n",
       "0                                                0.0     \n",
       "1                                                0.0     \n",
       "2                                                0.0     \n",
       "3                                                0.0     \n",
       "4                                                0.0     \n",
       "\n",
       "   specialRequest_QUANT_appendicular_lean_mass_residualizedRelated  \\\n",
       "0                                           0.000000                 \n",
       "1                                           0.000000                 \n",
       "2                                           0.000000                 \n",
       "3                                           0.000000                 \n",
       "4                                           0.132985                 \n",
       "\n",
       "   specialRequest_ageSquared_QUANT_age_squared  \\\n",
       "0                                         3969   \n",
       "1                                         2601   \n",
       "2                                         3364   \n",
       "3                                         4761   \n",
       "4                                         4096   \n",
       "\n",
       "   specialRequest_QUANT_years_observed_in_primary_HES  \\\n",
       "0                                           0.134247    \n",
       "1                                           0.002740    \n",
       "2                                           0.002740    \n",
       "3                                           0.002740    \n",
       "4                                           0.010959    \n",
       "\n",
       "   specialRequest_QUANT_OA_HES_based_age_of_onset  \n",
       "0                                             0.0  \n",
       "1                                             0.0  \n",
       "2                                             0.0  \n",
       "3                                             0.0  \n",
       "4                                             0.0  \n",
       "\n",
       "[5 rows x 638 columns]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "asthma_df_quant.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#data=asthma_df_quant.transpose().values\n",
    "asthma_df_quant_scaled=asthma_df_quant\n",
    "scaler = preprocessing.MinMaxScaler(feature_range=(0.0, 1.0))\n",
    "asthma_df_quant_scaled[asthma_df_quant_scaled.columns]=scaler.fit_transform(asthma_df_quant[asthma_df_quant.columns])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x7fd57146c210>"
      ]
     },
     "execution_count": 86,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(asthma_df['age'].values,asthma_df_quant_scaled['age'].values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "data_minmax=asthma_df_quant_scaled.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5000, 638)\n"
     ]
    }
   ],
   "source": [
    "print data_minmax.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "batch_size = 50\n",
    "original_dim = data.shape[1]\n",
    "#latent_dim = 2\n",
    "latent_dim=12\n",
    "#intermediate_dim = 500\n",
    "epochs = 200\n",
    "epsilon_std = 1.0\n",
    "learning_rate=0.0005\n",
    "momentum=0.9"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def sampling(args):\n",
    "    \n",
    "    import tensorflow as tf\n",
    "    epsilon_std=1.0\n",
    "    \n",
    "    z_mean, z_log_var = args\n",
    "    #epsilon = K.random_normal(shape=(batch_size, latent_dim), mean=0.,\n",
    "                             # stddev=epsilon_std)\n",
    "    epsilon = K.random_normal(shape=tf.shape(z_mean), mean=0.,\n",
    "                              stddev=epsilon_std)\n",
    "    return z_mean + K.exp(z_log_var / 2) * epsilon\n",
    "\n",
    "def vae_loss(x, x_decoded_mean):\n",
    "        xent_loss = original_dim * metrics.binary_crossentropy(x, x_decoded_mean)\n",
    "        kl_loss = - 0.5 * K.sum(1 + z_log_var_encoded - K.square(z_mean_encoded) - K.exp(z_log_var_encoded), axis=-1)\n",
    "        return K.mean(xent_loss + kl_loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Encoder\n",
    "x = Input(shape=(original_dim, ))\n",
    "\n",
    "# Input layer is compressed into a mean and log variance vector of size `latent_dim`\n",
    "# Each layer is initialized with glorot uniform weights and each step (dense connections, batch norm,\n",
    "# and relu activation) are funneled separately\n",
    "# Each vector of length `latent_dim` are connected to the input tensor\n",
    "# Use dropout to help with training and overfitting\n",
    "# Use batch normalization for regularization\n",
    "\n",
    "dropout_frac=0.1\n",
    "dropout_mean=Dropout(dropout_frac, input_shape=(original_dim,))(x)\n",
    "dropout_var=Dropout(dropout_frac, input_shape=(original_dim,))(x)\n",
    "\n",
    "z_mean_dense_linear = Dense(latent_dim, kernel_initializer='glorot_uniform')(dropout_mean)\n",
    "z_mean_dense_batchnorm = keras.layers.BatchNormalization(momentum=momentum)(z_mean_dense_linear)\n",
    "z_mean_encoded = keras.layers.Activation('relu')(z_mean_dense_batchnorm)\n",
    "\n",
    "z_log_var_dense_linear = Dense(latent_dim, kernel_initializer='glorot_uniform')(dropout_var)\n",
    "z_log_var_dense_batchnorm = keras.layers.BatchNormalization(momentum=momentum)(z_log_var_dense_linear)\n",
    "z_log_var_encoded = keras.layers.Activation('relu')(z_log_var_dense_batchnorm)\n",
    "\n",
    "# return the encoded and randomly sampled z vector\n",
    "# Takes two keras layers as input to the custom sampling function layer with a `latent_dim` output\n",
    "z = Lambda(sampling, output_shape=(latent_dim, ))([z_mean_encoded, z_log_var_encoded])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Decoder\n",
    "decoder = Dense(original_dim, kernel_initializer='glorot_uniform', activation='sigmoid')\n",
    "expression_reco = decoder(z)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "np.random.seed(123)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_1 (InputLayer)            (None, 638)          0                                            \n",
      "__________________________________________________________________________________________________\n",
      "dropout_1 (Dropout)             (None, 638)          0           input_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dropout_2 (Dropout)             (None, 638)          0           input_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_1 (Dense)                 (None, 12)           7668        dropout_1[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_2 (Dense)                 (None, 12)           7668        dropout_2[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_1 (BatchNor (None, 12)           48          dense_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_2 (BatchNor (None, 12)           48          dense_2[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "activation_1 (Activation)       (None, 12)           0           batch_normalization_1[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "activation_2 (Activation)       (None, 12)           0           batch_normalization_2[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "lambda_1 (Lambda)               (None, 12)           0           activation_1[0][0]               \n",
      "                                                                 activation_2[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "dense_3 (Dense)                 (None, 638)          8294        lambda_1[0][0]                   \n",
      "==================================================================================================\n",
      "Total params: 23,726\n",
      "Trainable params: 23,678\n",
      "Non-trainable params: 48\n",
      "__________________________________________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "adam = keras.optimizers.Adam(lr=learning_rate)\n",
    "\n",
    "#vae_layer = CustomVariationalLayer()([x, expression_reconstruct])\n",
    "#vae = Model(x, vae_layer)\n",
    "vae = Model(x, expression_reco)\n",
    "#RMS prop seems to be the most stable optimizer for this config\n",
    "vae.compile(optimizer=\"rmsprop\", loss=vae_loss)\n",
    "#vae.compile(optimizer=adam, loss=vae_loss)\n",
    "\n",
    "print vae.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import time\n",
    "start = time.clock()\n",
    "data_split=0.2\n",
    "data_train, data_test = train_test_split(data_minmax, test_size=data_split)\n",
    "tbCallBack = keras.callbacks.TensorBoard(log_dir='./Graph_AN_Variational', histogram_freq=0, write_graph=True, write_images=True)\n",
    "earlystop = keras.callbacks.EarlyStopping(monitor='loss', min_delta=0.0001, patience=5, verbose=1, mode='auto')\n",
    "callbacks_list = [tbCallBack,earlystop]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 3200 samples, validate on 800 samples\n",
      "Epoch 1/200\n",
      "3200/3200 [==============================] - 1s 228us/step - loss: 433.3987 - val_loss: 412.9438\n",
      "Epoch 2/200\n",
      "3200/3200 [==============================] - 0s 126us/step - loss: 403.9192 - val_loss: 384.5409\n",
      "Epoch 3/200\n",
      "3200/3200 [==============================] - 0s 131us/step - loss: 381.5680 - val_loss: 367.1635\n",
      "Epoch 4/200\n",
      "3200/3200 [==============================] - 0s 134us/step - loss: 364.4987 - val_loss: 355.5708\n",
      "Epoch 5/200\n",
      "3200/3200 [==============================] - 0s 135us/step - loss: 353.9066 - val_loss: 346.2848\n",
      "Epoch 6/200\n",
      "3200/3200 [==============================] - 0s 141us/step - loss: 347.1063 - val_loss: 343.9981\n",
      "Epoch 7/200\n",
      "3200/3200 [==============================] - 1s 215us/step - loss: 343.8761 - val_loss: 340.7625\n",
      "Epoch 8/200\n",
      "3200/3200 [==============================] - 1s 187us/step - loss: 341.6021 - val_loss: 340.1570\n",
      "Epoch 9/200\n",
      "3200/3200 [==============================] - 0s 130us/step - loss: 340.9425 - val_loss: 339.4549\n",
      "Epoch 10/200\n",
      "3200/3200 [==============================] - 0s 140us/step - loss: 339.9571 - val_loss: 338.5970\n",
      "Epoch 11/200\n",
      "3200/3200 [==============================] - 0s 132us/step - loss: 339.5994 - val_loss: 337.8890\n",
      "Epoch 12/200\n",
      "3200/3200 [==============================] - 0s 136us/step - loss: 338.6870 - val_loss: 337.7622\n",
      "Epoch 13/200\n",
      "3200/3200 [==============================] - 0s 141us/step - loss: 338.7596 - val_loss: 337.9452\n",
      "Epoch 14/200\n",
      "3200/3200 [==============================] - 0s 130us/step - loss: 338.3228 - val_loss: 337.0286\n",
      "Epoch 15/200\n",
      "3200/3200 [==============================] - 0s 131us/step - loss: 338.1320 - val_loss: 337.2338\n",
      "Epoch 16/200\n",
      "3200/3200 [==============================] - 0s 140us/step - loss: 337.6602 - val_loss: 336.9779\n",
      "Epoch 17/200\n",
      "3200/3200 [==============================] - 0s 136us/step - loss: 337.6262 - val_loss: 336.3813\n",
      "Epoch 18/200\n",
      "3200/3200 [==============================] - 0s 135us/step - loss: 337.5405 - val_loss: 336.0552\n",
      "Epoch 19/200\n",
      "3200/3200 [==============================] - 0s 139us/step - loss: 337.0055 - val_loss: 336.4154\n",
      "Epoch 20/200\n",
      "3200/3200 [==============================] - 0s 147us/step - loss: 336.9888 - val_loss: 336.7719\n",
      "Epoch 21/200\n",
      "3200/3200 [==============================] - 0s 147us/step - loss: 336.6019 - val_loss: 336.1969\n",
      "Epoch 22/200\n",
      "3200/3200 [==============================] - 0s 135us/step - loss: 336.9548 - val_loss: 335.6005\n",
      "Epoch 23/200\n",
      "3200/3200 [==============================] - 0s 142us/step - loss: 336.6377 - val_loss: 335.2109\n",
      "Epoch 24/200\n",
      "3200/3200 [==============================] - 0s 129us/step - loss: 336.5824 - val_loss: 336.0532\n",
      "Epoch 25/200\n",
      "3200/3200 [==============================] - 0s 138us/step - loss: 336.2525 - val_loss: 335.7146\n",
      "Epoch 26/200\n",
      "3200/3200 [==============================] - 0s 133us/step - loss: 336.1169 - val_loss: 335.9279\n",
      "Epoch 27/200\n",
      "3200/3200 [==============================] - 0s 134us/step - loss: 335.9972 - val_loss: 335.6663\n",
      "Epoch 28/200\n",
      "3200/3200 [==============================] - 0s 137us/step - loss: 336.1159 - val_loss: 336.7820\n",
      "Epoch 29/200\n",
      "3200/3200 [==============================] - 1s 165us/step - loss: 336.0305 - val_loss: 335.3935\n",
      "Epoch 30/200\n",
      "3200/3200 [==============================] - 0s 143us/step - loss: 335.7204 - val_loss: 335.5568\n",
      "Epoch 31/200\n",
      "3200/3200 [==============================] - 0s 130us/step - loss: 335.8354 - val_loss: 336.5758\n",
      "Epoch 32/200\n",
      "3200/3200 [==============================] - 0s 142us/step - loss: 335.6236 - val_loss: 335.1047\n",
      "Epoch 33/200\n",
      "3200/3200 [==============================] - 0s 139us/step - loss: 336.0668 - val_loss: 335.0433\n",
      "Epoch 34/200\n",
      "3200/3200 [==============================] - 0s 132us/step - loss: 335.5491 - val_loss: 335.0267\n",
      "Epoch 35/200\n",
      "3200/3200 [==============================] - 0s 137us/step - loss: 335.5373 - val_loss: 334.5763\n",
      "Epoch 36/200\n",
      "3200/3200 [==============================] - 0s 141us/step - loss: 335.4429 - val_loss: 334.3880\n",
      "Epoch 37/200\n",
      "3200/3200 [==============================] - 0s 135us/step - loss: 335.6092 - val_loss: 335.3000\n",
      "Epoch 38/200\n",
      "3200/3200 [==============================] - 0s 136us/step - loss: 335.5210 - val_loss: 334.8898\n",
      "Epoch 39/200\n",
      "3200/3200 [==============================] - 0s 130us/step - loss: 335.1994 - val_loss: 334.1404\n",
      "Epoch 40/200\n",
      "3200/3200 [==============================] - 0s 135us/step - loss: 335.0057 - val_loss: 335.0195\n",
      "Epoch 41/200\n",
      "3200/3200 [==============================] - 0s 137us/step - loss: 335.0215 - val_loss: 334.7983\n",
      "Epoch 42/200\n",
      "3200/3200 [==============================] - 0s 140us/step - loss: 335.3975 - val_loss: 335.0619\n",
      "Epoch 43/200\n",
      "3200/3200 [==============================] - 0s 130us/step - loss: 335.2182 - val_loss: 334.6054\n",
      "Epoch 44/200\n",
      "3200/3200 [==============================] - 0s 139us/step - loss: 334.9120 - val_loss: 335.0387\n",
      "Epoch 45/200\n",
      "3200/3200 [==============================] - 0s 137us/step - loss: 334.8196 - val_loss: 334.6822\n",
      "Epoch 46/200\n",
      "3200/3200 [==============================] - 0s 135us/step - loss: 335.0931 - val_loss: 333.8532\n",
      "Epoch 47/200\n",
      "3200/3200 [==============================] - 0s 137us/step - loss: 335.0813 - val_loss: 335.1864\n",
      "Epoch 48/200\n",
      "3200/3200 [==============================] - 0s 135us/step - loss: 334.9783 - val_loss: 334.6629\n",
      "Epoch 49/200\n",
      "3200/3200 [==============================] - 0s 132us/step - loss: 334.9155 - val_loss: 335.0168\n",
      "Epoch 50/200\n",
      "3200/3200 [==============================] - 0s 135us/step - loss: 334.6635 - val_loss: 335.0479\n",
      "Epoch 51/200\n",
      "3200/3200 [==============================] - 0s 130us/step - loss: 334.7250 - val_loss: 335.4248\n",
      "Epoch 52/200\n",
      "3200/3200 [==============================] - 0s 147us/step - loss: 334.3859 - val_loss: 334.4000\n",
      "Epoch 53/200\n",
      "3200/3200 [==============================] - 0s 136us/step - loss: 334.7946 - val_loss: 334.5679\n",
      "Epoch 54/200\n",
      "3200/3200 [==============================] - 0s 140us/step - loss: 334.4332 - val_loss: 334.6635\n",
      "Epoch 55/200\n",
      "3200/3200 [==============================] - 0s 139us/step - loss: 334.6181 - val_loss: 334.6957\n",
      "Epoch 56/200\n",
      "3200/3200 [==============================] - 0s 141us/step - loss: 334.5283 - val_loss: 335.2469\n",
      "Epoch 57/200\n",
      "3200/3200 [==============================] - 0s 137us/step - loss: 334.5825 - val_loss: 335.7194\n",
      "Epoch 00057: early stopping\n"
     ]
    }
   ],
   "source": [
    "hist1=vae.fit(data_train,data_train,\n",
    "        #shuffle=True,\n",
    "        epochs=epochs,\n",
    "        batch_size=batch_size,\n",
    "        validation_split=0.2,\n",
    "        callbacks=callbacks_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Visualize training performance\n",
    "history_df = pd.DataFrame(hist1.history)\n",
    "ax = history_df[1:].plot()\n",
    "ax.set_xlabel('Epochs')\n",
    "ax.set_ylabel('VAE Loss')\n",
    "fig = ax.get_figure()\n",
    "#plt.yscale('log')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Time to train:  106.76\n"
     ]
    }
   ],
   "source": [
    "trained = time.clock()\n",
    "print \"Time to train: \",trained-start"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# encoder, from inputs to latent space\n",
    "encoder = Model(x, z_mean_encoded)\n",
    "#encoder = Model(x, z)\n",
    "\n",
    "# generator, from latent space to reconstructed inputs\n",
    "decoder_input = Input(shape=(latent_dim,))\n",
    "decoded_latent=decoder(decoder_input)\n",
    "generator=Model(decoder_input,decoded_latent)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.0002713506199267008 0.11542296422686471\n"
     ]
    }
   ],
   "source": [
    "#data_predict=vae.predict(data_test,batch_size=batch_size)\n",
    "data_predict=vae.predict_on_batch(data_test)\n",
    "diffs=np.subtract(data_predict,data_test).flatten()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.0002713506199267008 0.11542296422686471\n"
     ]
    }
   ],
   "source": [
    "plt.hist(diffs, bins='auto')  # plt.hist passes it's arguments to np.histogram\n",
    "\n",
    "plt.title(\"Predicted phenotype minus real phenotype (normalized)\")\n",
    "plt.xlim(-0.3,0.3)\n",
    "plt.show()\n",
    "print np.mean(diffs),np.std(diffs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "encoded_predict=encoder.predict(data_test)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "tsne = manifold.TSNE(n_components=2, init='pca', random_state=0, perplexity=50,\n",
    "                     learning_rate=10, n_iter=1000)\n",
    "#tsne = manifold.TSNE(n_components=2, init='pca')\n",
    "\n",
    "tsne_out = tsne.fit_transform(encoded_predict)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(tsne_out[:,0],tsne_out[:,1])\n",
    "\n",
    "plt.xlabel('Dimension 1')\n",
    "plt.ylabel('Dimension 2')\n",
    "plt.title('Autoencoder QUANTITY FEATURES ONLY tSNE with two components')\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn import decomposition\n",
    "\n",
    "pca = decomposition.PCA(n_components=2)\n",
    "pca_out=pca.fit_transform(encoded_predict)\n",
    "\n",
    "plt.scatter(pca_out[:,0],pca_out[:,1])\n",
    "\n",
    "plt.xlabel('Principal Component 1')\n",
    "plt.ylabel('Principal Component 2')\n",
    "plt.title('Autoencoder QUANTITY FEATURES ONLY')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:anaconda-default]",
   "language": "python",
   "name": "conda-env-anaconda-default-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.14"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}