dropout masking .ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import sklearn \n",
    "import scipy\n",
    "from sklearn import linear_model as lm\n",
    "from sklearn.linear_model import LinearRegression\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.cross_validation import KFold, train_test_split, cross_val_score, StratifiedKFold, LabelKFold, ShuffleSplit\n",
    "from sklearn.metrics import roc_auc_score\n",
    "from sklearn.svm import SVR\n",
    "from sklearn.ensemble import GradientBoostingRegressor\n",
    "from mhcflurry.amino_acid import common_amino_acids\n",
    "from mhcflurry import dataset\n",
    "from mhcflurry.dataset import Dataset\n",
    "import matplotlib.pyplot as plt \n",
    "% matplotlib inline\n",
    "import numpy as np\n",
    "import math \n",
    "from mhcflurry import peptide_encoding, amino_acid\n",
    "import statsmodels.api as sm\n",
    "from keras import models, layers, optimizers\n",
    "from keras.optimizers import Adam \n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Embedding, LSTM, Input, merge, Convolution1D, AveragePooling1D, Activation, Flatten\n",
    "from keras.preprocessing import sequence\n",
    "from keras.models import Model\n",
    "from keras.engine import topology\n",
    "import seaborn as sns\n",
    "from keras.layers.core import Lambda\n",
    "from keras import backend as K"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "ds = Dataset.from_csv(\"bdata.2009.mhci.public.1.txt\")\n",
    "ds_h = ds.slice(ds.alleles == 'HLA-A0201')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "df = pd.read_table(\"bdata.2009.mhci.public.1.txt\")\n",
    "\n",
    "df['log_meas']=1-np.log(df['meas'])/math.log(50000)\n",
    "df['peptide_length'] = df['sequence'].str.len()\n",
    "\n",
    "\n",
    "max_len=df['sequence'].str.len().max()\n",
    "n_peptides = df['sequence'].count()\n",
    "\n",
    "def amino_acid_hotshot_encoding(s):\n",
    "    return common_amino_acids.hotshot_encoding([s],len(s)).flatten().astype(int)\n",
    "df['hotshot_encoded_peptides'] = df.sequence.apply(lambda seq: amino_acid_hotshot_encoding(seq))\n",
    "\n",
    "def amino_acid_index_encoding(s, maxlen):\n",
    "    a = 1+common_amino_acids.index_encoding([s],len(s)).flatten()\n",
    "    return np.concatenate([a, np.zeros(maxlen-len(a),dtype=int)])\n",
    "df['index_encoded_peptides'] = df.sequence.apply(lambda seq: amino_acid_index_encoding(seq, max_len))\n",
    "\n",
    "def measured_affinity_less_than(Y,k):\n",
    "    IC50 = 50000**(1-Y)\n",
    "    return (IC50 < k).astype(int) \n",
    "\n",
    "def affinity_label(Y):\n",
    "    return measured_affinity_less_than(Y,50) + measured_affinity_less_than(Y,500) + measured_affinity_less_than(Y,5000) + measured_affinity_less_than(Y,50000)\n",
    "\n",
    "df['affinity_label'] = affinity_label(df['log_meas'])\n",
    "df_h = df[df['mhc']=='HLA-A-0201'][['hotshot_encoded_peptides','index_encoded_peptides','log_meas','peptide_length']]\n",
    "X = np.array(list(df_h['index_encoded_peptides']))\n",
    "y = np.array(list(df_h['log_meas']))\n",
    "y[y<0]=0\n",
    "\n",
    "def first_and_last_three(Y):\n",
    "    k = np.count_nonzero(Y)\n",
    "    return np.concatenate([Y[:3],Y[-3+k:k]])\n",
    "def first_and_last_four(Y):\n",
    "    k = np.count_nonzero(Y)\n",
    "    return np.concatenate([Y[:4],Y[-4+k:k]])\n",
    "def first_and_last_two(Y):\n",
    "    k = np.count_nonzero(Y)\n",
    "    return np.concatenate([Y[:2],Y[-2+k:k]])\n",
    "X_44 = np.apply_along_axis(first_and_last_four,1,X)\n",
    "X_33 = np.apply_along_axis(first_and_last_three,1,X)\n",
    "X_22 = np.apply_along_axis(first_and_last_two,1,X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def regroup_together(affinities, weights , original_indices):\n",
    "    affinities = affinities.ravel()\n",
    "    weights = weights.ravel()\n",
    "    \n",
    "    assert affinities.shape == weights.shape, \"%s should be %s\" % (affinities.shape, weights.shape)\n",
    "    assert affinities.shape == original_indices.shape\n",
    "    assert len(affinities) == len(affinities.ravel())\n",
    "    \n",
    "    weighted_affinities = (affinities * weights)\n",
    "    index_set = set(original_indices)\n",
    "    n_indices = len(index_set)\n",
    "    result_order = {original_index: i for (i, original_index) in enumerate(sorted(index_set))}\n",
    "    result = np.zeros(n_indices)\n",
    "    for i, x in enumerate(weighted_affinities):\n",
    "        result_idx = result_order[original_indices[i]]\n",
    "        result[result_idx] += x\n",
    "    return result\n",
    "\n",
    "def slicing(dataset, index, i):\n",
    "    return dataset.slice(index).kmer_index_encoding()[i]\n",
    "\n",
    "def label_transform(array):\n",
    "    result = 1-np.log(array)/math.log(50000)\n",
    "    result[result<0]=0\n",
    "    return result\n",
    "\n",
    "def index_to_hotshot_encoding(index_encoded_nine_mer):\n",
    "    result = np.zeros((9,21))\n",
    "    for position, amino_acid in enumerate(index_encoded_nine_mer):\n",
    "        result[position][amino_acid]= 1\n",
    "    return result.flatten()\n",
    "\n",
    "def real_labels(dataset,index):\n",
    "    \n",
    "    y = label_transform(slicing(dataset,index,1))\n",
    "    weights = slicing(dataset,index,2)\n",
    "    original_indices = slicing(dataset,index,3)\n",
    "    \n",
    "    return regroup_together(y, weights , original_indices)\n",
    "\n",
    "def fit(model,dataset,index, neural_network = False, hotshot = False): # to be left out or modified \n",
    "    \n",
    "    X = slicing(dataset,index,0)\n",
    "    \n",
    "    if (hotshot == True):\n",
    "        X = np.apply_along_axis(index_to_hotshot_encoding, 1, X)\n",
    "        \n",
    "    y = label_transform(slicing(dataset,index,1))\n",
    "    weights = slicing(dataset,index,2)\n",
    "    \n",
    "    if (neural_network == True):\n",
    "        model.fit(X, y, sample_weight = weights, batch_size = 16, nb_epoch = 1)\n",
    "    else: \n",
    "        model.fit(X, y, sample_weight = weights)\n",
    "        \n",
    "def predict(model, dataset, index, hotshot = False):\n",
    "    \n",
    "    X = slicing(dataset,index,0)\n",
    "    \n",
    "    if (hotshot == True):\n",
    "        X = np.apply_along_axis(index_to_hotshot_encoding, 1, X)\n",
    "        \n",
    "    weights = slicing(dataset,index,2)\n",
    "    original_indices = slicing(dataset,index,3)\n",
    "    \n",
    "    return regroup_together(model.predict(X), weights , original_indices)\n",
    "\n",
    "def AUC(model, dataset, index, hotshot = False):\n",
    "        \n",
    "    real_affinity = measured_affinity_less_than(real_labels(dataset,index),500)\n",
    "    predicted_affinity = predict(model, dataset, index, hotshot = hotshot)\n",
    "    \n",
    "    return roc_auc_score(real_affinity, predicted_affinity)\n",
    "\n",
    "def AUC_simple(model, features, labels, index):\n",
    "    real_affinity = measured_affinity_less_than(labels[index],500)\n",
    "    predicted_affinity = model.predict(features[index])\n",
    "    \n",
    "    return roc_auc_score(real_affinity, predicted_affinity)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def split_by_length(X,index,length=9):\n",
    "    length_idx = np.array([i for i in index if (np.count_nonzero(X[i])==length)])\n",
    "    non_length_idx = np.array([i for i in index if (np.count_nonzero(X[i])!=length)])\n",
    "    return index, length_idx, non_length_idx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def random_dropout_prediction_by_lentgh(model,array,length):\n",
    "    array_of_lengths = np.apply_along_axis(np.count_nonzero,1,array)\n",
    "    #print(\"array shape\", array.shape, \"array of lengths shape\", array_of_lengths.shape)\n",
    "    bool_array = (array_of_lengths == length)\n",
    "    #print(\"bool_array shape\", bool_array.shape)\n",
    "    result = np.zeros(len(array))\n",
    "    #print(\"result shape\", result.shape)\n",
    "    for i in range(length):\n",
    "        result[bool_array] = result[bool_array] + model.predict(array[bool_array])[:,0]\n",
    "    #print(\"result[bool_array] shape\", result[bool_array].shape, \"model prediction shape\", model.predict(array[bool_array]).shape)\n",
    "    return result/length\n",
    "\n",
    "def random_dropout_array_prediction(model,array):\n",
    "    array_of_lengths = np.apply_along_axis(np.count_nonzero,1,array)\n",
    "    result = np.zeros(len(array))\n",
    "    for length in np.unique(array_of_lengths):\n",
    "        result = result + random_dropout_prediction_by_lentgh(model,array,length)\n",
    "    return result\n",
    "\n",
    "def AUC_random_dropout(model, features, labels, index):\n",
    "    real_affinity = measured_affinity_less_than(labels[index],500)\n",
    "    predicted_affinity = random_dropout_array_prediction(model, features[index])\n",
    "    \n",
    "    return roc_auc_score(real_affinity, predicted_affinity)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 12s - loss: 0.0436    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 12s - loss: 0.0444    \n",
      "allmers: 0.921899185738 0.925555039398 0 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 12s - loss: 0.0281    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 12s - loss: 0.0279    \n",
      "allmers: 0.946171477154 0.943995144227 0 1\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 12s - loss: 0.0249    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0253    \n",
      "allmers: 0.948186097063 0.947503308762 0 2\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 12s - loss: 0.0236    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0241    \n",
      "allmers: 0.949711633696 0.948814542777 0 3\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0228    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 14s - loss: 0.0232    \n",
      "allmers: 0.948947331516 0.949561753337 0 4\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0222    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0227    \n",
      "allmers: 0.949444741478 0.949633187543 0 5\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0220    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0229    \n",
      "allmers: 0.949807609715 0.950012270907 0 6\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 14s - loss: 0.0221    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 14s - loss: 0.0224    \n",
      "allmers: 0.949909282941 0.950085458099 0 7\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 14s - loss: 0.0219    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0220    \n",
      "allmers: 0.949843545941 0.950215617358 0 8\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0215     \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0220     \n",
      "allmers: 0.950084143359 0.950160836525 0 9\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0216     \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0219     \n",
      "allmers: 0.949853187368 0.95034928259 0 10\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0216     \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0217     \n",
      "allmers: 0.950628883961 0.950305896171 0 11\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0212     \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 16s - loss: 0.0217    \n",
      "allmers: 0.949791394589 0.950236214951 0 12\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 10s - loss: 0.0213    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 10s - loss: 0.0216    \n",
      "allmers: 0.949957051827 0.950226573525 0 13\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0212     \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0217     \n",
      "allmers: 0.949900079761 0.950269083451 0 14\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 11s - loss: 0.0212    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0218     \n",
      "allmers: 0.950090278813 0.950275657151 0 15\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0215     \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0214     \n",
      "allmers: 0.950184501845 0.950247609364 0 16\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0213    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 18s - loss: 0.0216    \n",
      "allmers: 0.949899641514 0.950278286631 0 17\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 14s - loss: 0.0208    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 16s - loss: 0.0214    \n",
      "allmers: 0.950386533557 0.950312469871 0 18\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0214     \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0216     \n",
      "allmers: 0.950853704499 0.950294063511 0 19\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0211     \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0215     \n",
      "allmers: 0.950726612967 0.950299322471 0 20\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0210     \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0216     \n",
      "allmers: 0.950114382379 0.950288366304 0 21\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0213     \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0214     \n",
      "allmers: 0.950127091532 0.950333286587 0 22\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0209     \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 11s - loss: 0.0217    \n",
      "allmers: 0.94999211156 0.95032824675 0 23\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 18s - loss: 0.0209    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 16s - loss: 0.0215    \n",
      "allmers: 0.950543425862 0.95031904357 0 24\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 15s - loss: 0.0209    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 16s - loss: 0.0214    \n",
      "allmers: 0.950565338195 0.950327370257 0 25\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 15s - loss: 0.0208    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 14s - loss: 0.0211    \n",
      "allmers: 0.950462788476 0.950351473824 0 26\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0209    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 11s - loss: 0.0215    \n",
      "allmers: 0.950502230676 0.950344900124 0 27\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 11s - loss: 0.0211    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0214     \n",
      "allmers: 0.950268206958 0.950339202917 0 28\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0210     \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0214     \n",
      "allmers: 0.950322549544 0.95032693201 0 29\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0209     \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0213     \n",
      "allmers: 0.949898326774 0.950309402144 0 30\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0212    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 10s - loss: 0.0213    \n",
      "allmers: 0.95071434206 0.95032035831 0 31\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 11s - loss: 0.0210    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0213     \n",
      "allmers: 0.950029362526 0.950315099351 0 32\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0211     \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0214     \n",
      "allmers: 0.950514063335 0.950301951951 0 33\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 11s - loss: 0.0209    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 16s - loss: 0.0213    \n",
      "allmers: 0.950335696944 0.950323864284 0 34\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 15s - loss: 0.0210    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0215    \n",
      "allmers: 0.950862031186 0.950322549544 0 35\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 14s - loss: 0.0211    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 14s - loss: 0.0214    \n",
      "allmers: 0.950940477338 0.95032561727 0 36\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 14s - loss: 0.0210    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 10s - loss: 0.0215    \n",
      "allmers: 0.950966772138 0.95032298779 0 37\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0208    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 12s - loss: 0.0214    \n",
      "allmers: 0.950351035577 0.950308525651 0 38\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 12s - loss: 0.0210    \n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0212     \n",
      "allmers: 0.950710836087 0.950305896171 0 39\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 9s - loss: 0.0435     \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 18s - loss: 0.0456    \n",
      "allmers: 0.934665928088 0.924255566364 1 0\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0282     \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0289     \n",
      "allmers: 0.944948389537 0.946592853797 1 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0249     \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0251     \n",
      "allmers: 0.947949992794 0.948861282245 1 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0239     \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0241     \n",
      "allmers: 0.948548962387 0.949560455397 1 3\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0235     \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0234     \n",
      "allmers: 0.948938968151 0.949135772446 1 4\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0227     \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0230     \n",
      "allmers: 0.948790801989 0.949583423404 1 5\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0224     \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 9s - loss: 0.0223     \n",
      "allmers: 0.949468133016 0.94956090575 1 6\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0222     \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0221    \n",
      "allmers: 0.949456423836 0.949514969736 1 7\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 9s - loss: 0.0221     \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 9s - loss: 0.0220     \n",
      "allmers: 0.949442913244 0.949671242254 1 8\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0220     \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0219     \n",
      "allmers: 0.949013276409 0.949609994235 1 9\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0218    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0218    \n",
      "allmers: 0.949436608301 0.949504161262 1 10\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0218    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0217    \n",
      "allmers: 0.949056960657 0.949564508575 1 11\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0217    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0219    \n",
      "allmers: 0.949162343277 0.94960188788 1 12\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0218    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0215    \n",
      "allmers: 0.94938526805 0.94965457919 1 13\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0215    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0218    \n",
      "allmers: 0.949151985156 0.949611795648 1 14\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0217    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0215    \n",
      "allmers: 0.949357796512 0.94960188788 1 15\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0216    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0211    \n",
      "allmers: 0.949312761205 0.949557753279 1 16\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0215    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0216    \n",
      "allmers: 0.949459576308 0.94954919657 1 17\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0217    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0213    \n",
      "allmers: 0.948991209108 0.949554150454 1 18\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0217    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0212    \n",
      "allmers: 0.949159641159 0.949550097276 1 19\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0211    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0213    \n",
      "allmers: 0.949334378152 0.949545143392 1 20\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0213    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0211    \n",
      "allmers: 0.949050205361 0.949547845511 1 21\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0216    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0214    \n",
      "allmers: 0.949377161695 0.949547845511 1 22\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0213    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0213    \n",
      "allmers: 0.948980850987 0.94954334198 1 23\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0216    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 1222s - loss: 0.0210  \n",
      "allmers: 0.948990308402 0.94954334198 1 24\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0211    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0210    \n",
      "allmers: 0.949120010088 0.949557753279 1 25\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0214    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 15s - loss: 0.0213    \n",
      "allmers: 0.948973194985 0.949570813518 1 26\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0213    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0213    \n",
      "allmers: 0.949564958928 0.949569912812 1 27\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0215    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0210    \n",
      "allmers: 0.949544242686 0.949567661046 1 28\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0212    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0211    \n",
      "allmers: 0.949733390979 0.949566309987 1 29\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0214    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0211    \n",
      "allmers: 0.949493352789 0.949582522698 1 30\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0211    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0212    \n",
      "allmers: 0.949634763655 0.949561806456 1 31\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0214    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0212    \n",
      "allmers: 0.949164144689 0.949580270932 1 32\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 13s - loss: 0.0212    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 14s - loss: 0.0210    \n",
      "allmers: 0.949562932339 0.949572164577 1 33\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 13s - loss: 0.0214    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 14s - loss: 0.0212    \n",
      "allmers: 0.94919206658 0.949580721285 1 34\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 14s - loss: 0.0213    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 14s - loss: 0.0211    \n",
      "allmers: 0.949426700533 0.94957846952 1 35\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 3210s - loss: 0.0211  \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 12s - loss: 0.0211    \n",
      "allmers: 0.949599636115 0.949582522698 1 36\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0212    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0209    \n",
      "allmers: 0.949148832685 0.949583423404 1 37\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0215    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0214    \n",
      "allmers: 0.949106049143 0.949581621992 1 38\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0212    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0212    \n",
      "allmers: 0.949407785704 0.949580721285 1 39\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0427    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0446    \n",
      "allmers: 0.92518528476 0.931833579782 2 0\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0284    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0273    \n",
      "allmers: 0.943577093695 0.943947663216 2 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0249    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0248    \n",
      "allmers: 0.943923018327 0.94619258861 2 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0237    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0237    \n",
      "allmers: 0.945478334902 0.946796612448 2 3\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0230    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 14s - loss: 0.0230    \n",
      "allmers: 0.944925841287 0.947108930412 2 4\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 12s - loss: 0.0227    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0224    \n",
      "allmers: 0.945660258995 0.947259936371 2 5\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0220    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 2200s - loss: 0.0220  \n",
      "allmers: 0.94578482771 0.947285477439 2 6\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 7905s - loss: 0.0221   \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 14s - loss: 0.0218    \n",
      "allmers: 0.945489537124 0.947526997356 2 7\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 14s - loss: 0.0220    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 15s - loss: 0.0222    \n",
      "allmers: 0.945872653134 0.947622440292 2 8\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0218    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0215    \n",
      "allmers: 0.945009633911 0.947665456827 2 9\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0217    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0214    \n",
      "allmers: 0.945944347359 0.947577631402 2 10\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0214    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0216    \n",
      "allmers: 0.945476542546 0.947583008469 2 11\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0212    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0217    \n",
      "allmers: 0.945887888157 0.947667249182 2 12\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0213    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0215    \n",
      "allmers: 0.945934489403 0.947637675315 2 13\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0217    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0212    \n",
      "allmers: 0.946012008783 0.947592866425 2 14\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0213    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0213    \n",
      "allmers: 0.945472509746 0.947648429448 2 15\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0211    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0215    \n",
      "allmers: 0.945508804947 0.947675314782 2 16\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0211    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0213    \n",
      "allmers: 0.946274588878 0.947622440292 2 17\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 21s - loss: 0.0213    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0212    \n",
      "allmers: 0.945692073307 0.94766007976 2 18\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 13s - loss: 0.0210    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0211    \n",
      "allmers: 0.945575122104 0.947653358426 2 19\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0213    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0215    \n",
      "allmers: 0.94574001882 0.947674866694 2 20\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0211    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0211    \n",
      "allmers: 0.945512389658 0.947694134516 2 21\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0211    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0211    \n",
      "allmers: 0.94608639154 0.947718779406 2 22\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0212    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0213    \n",
      "allmers: 0.946031276605 0.947703544383 2 23\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0211    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0212    \n",
      "allmers: 0.945899986557 0.947765380652 2 24\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0210    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0213    \n",
      "allmers: 0.945867276068 0.947763140207 2 25\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0210    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 12s - loss: 0.0213    \n",
      "allmers: 0.94598467536 0.947760451674 2 26\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0212    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0212    \n",
      "allmers: 0.945743827575 0.947784648474 2 27\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0215    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0210    \n",
      "allmers: 0.945813953488 0.947780167585 2 28\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0210    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0212    \n",
      "allmers: 0.94592328718 0.947785992741 2 29\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0211    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0212    \n",
      "allmers: 0.946016489672 0.947768517274 2 30\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0210    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0210    \n",
      "allmers: 0.946266523278 0.947780167585 2 31\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0209    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0210    \n",
      "allmers: 0.945753013398 0.947779719496 2 32\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0210    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0213    \n",
      "allmers: 0.945667876507 0.94780660483 2 33\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0212    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0211    \n",
      "allmers: 0.945926423802 0.94780660483 2 34\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0211    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0212    \n",
      "allmers: 0.946214544966 0.947811533808 2 35\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0209    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0210    \n",
      "allmers: 0.945512389658 0.947812878075 2 36\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0209    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0210    \n",
      "allmers: 0.94559304566 0.947814670431 2 37\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 801s - loss: 0.0210   \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0212    \n",
      "allmers: 0.945998118027 0.947822287942 2 38\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0209    \n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0210    \n",
      "allmers: 0.946176905498 0.94782318412 2 39\n"
     ]
    }
   ],
   "source": [
    "folds = 3\n",
    "batch_size_nn = 16\n",
    "batch_size_lstm = 16\n",
    "hidden = 50\n",
    "dropout_probability = 0.25\n",
    "\n",
    "n_epochs = 40\n",
    "epoch = 0\n",
    "\n",
    "\n",
    "\n",
    "lstm_avg_aucs = np.zeros((6, folds,n_epochs))\n",
    "lstm_aucs = np.zeros((6, folds,n_epochs))\n",
    "\n",
    "\n",
    "for i, (train_idx, test_idx) in enumerate(KFold(len(df_h),folds, shuffle=True)):\n",
    "    \n",
    "    list_index = split_by_length(X,train_idx,length=9)+split_by_length(X,test_idx,length=9)\n",
    "    \n",
    "    \n",
    "    # lstm_avg \n",
    "    sequence = Input( shape= (26, ),dtype='int32')\n",
    "    embedded = Embedding(input_dim = 21, output_dim= 32, mask_zero = True)(sequence)\n",
    "    forwards = LSTM(hidden)(embedded)\n",
    "    backwards = LSTM(hidden, go_backwards=True)(embedded)\n",
    "\n",
    "    merged = merge([forwards, backwards], mode = 'concat', concat_axis=-1)\n",
    "    after_dp = Lambda(lambda x: K.dropout(x, level=0.25))(merged)\n",
    "    output = Dense(1, activation = 'sigmoid')(after_dp)\n",
    "    lstm_avg = Model(input = sequence, output = output)\n",
    "    adam_avg = Adam(lr = 0.01)\n",
    "    lstm_avg.compile(optimizer = adam_avg, loss='mean_squared_error')\n",
    "    \n",
    "    # lstm \n",
    "    sequence = Input( shape= (26, ),dtype='int32')\n",
    "    embedded = Embedding(input_dim = 21, output_dim= 32, mask_zero = True)(sequence)\n",
    "    forwards = LSTM(hidden)(embedded)\n",
    "    backwards = LSTM(hidden, go_backwards=True)(embedded)\n",
    "\n",
    "    merged = merge([forwards, backwards], mode = 'concat', concat_axis=-1)\n",
    "    after_dp = Dropout(dropout_probability)(merged)\n",
    "    output = Dense(1, activation = 'sigmoid')(after_dp)\n",
    "    lstm = Model(input = sequence, output = output)\n",
    "    adam = Adam(lr = 0.01)\n",
    "    lstm.compile(optimizer = adam, loss='mean_squared_error')\n",
    "      \n",
    "    \n",
    "    for epoch in range(n_epochs):\n",
    "        \n",
    "        \n",
    "        #lstm_avg \n",
    "        adam_avg.lr.set_value(0.01*(1+epoch)**(-2))\n",
    "        lstm_avg.fit(X[train_idx],y[train_idx], batch_size = 16, nb_epoch = 1)\n",
    "        for k, index in enumerate(list_index):\n",
    "            lstm_avg_aucs[k][i][epoch] = AUC_random_dropout(lstm_avg, X, y, index)\n",
    "            \n",
    "        # lstm    \n",
    "        adam.lr.set_value(0.01*(1+epoch)**(-2))\n",
    "        lstm.fit(X[train_idx],y[train_idx], batch_size = 16, nb_epoch = 1)\n",
    "        for k, index in enumerate(list_index):\n",
    "            lstm_aucs[k][i][epoch] = AUC_random_dropout(lstm, X, y, index)\n",
    "        \n",
    "        print(\"allmers:\", lstm_avg_aucs[3][i][epoch], lstm_aucs[3][i][epoch], i, epoch)  \n",
    "\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x13d19df28>"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAArQAAAHcCAYAAAAncTlwAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\na/8Hn2t6iRzt7qcAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x13e949a20>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot( np.arange(0,40,1), lstm_aucs[3,0,0:40],color='r',marker='*', linestyle='-', label =\"test LSTM \")\n",
    "plt.plot( np.arange(0,40,1), lstm_avg_aucs[3,0,0:40],color='g',marker='*', linestyle='-', label =\"test LSTM AVG\")\n",
    "\n",
    "plt.legend(loc=4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 1,  1,  1, 16, 16, 17,  7, 15,  9, 18,  0,  0,  0,  0,  0,  0,  0,\n",
       "        0,  0,  0,  0,  0,  0,  0,  0,  0])"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a[:,None][:,0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 1,  1,  1, 16, 16, 17,  7, 15,  9, 18,  0,  0,  0,  0,  0,  0,  0,\n",
       "        0,  0,  0,  0,  0,  0,  0,  0,  0])"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.ravel(a[:,None])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "a=X[0:20]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 1,  5,  4,  5,  8, 12, 16, 10, 10,  9,  0,  0,  0,  0,  0,  0,  0,\n",
       "        0,  0,  0,  0,  0,  0,  0,  0,  0])"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X[20]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2356"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(np.apply_along_axis(np.count_nonzero,1,X) == 10).sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "8\n",
      "9\n",
      "10\n",
      "11\n",
      "12\n",
      "13\n",
      "14\n",
      "15\n"
     ]
    }
   ],
   "source": [
    "for i in np.unique(np.apply_along_axis(np.count_nonzero,1,X)):\n",
    "    print(i)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [conda root]",
   "language": "python",
   "name": "conda-root-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}