linear vs linear22 vs linear33 vs linear44 vs LSTM.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 15,
   "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, f1_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.models import Sequential\n",
    "from keras.optimizers import Adam, SGD\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"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "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": 17,
   "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"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def regroup_together(affinities, weights , original_indices):\n",
    "    affinities = affinities.ravel()\n",
    "    weights = weights.ravel()\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",
    "    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",
    "def slicing(dataset,index, i):\n",
    "    return dataset.slice(index).kmer_index_encoding()[i]\n",
    "def label_transform(array):\n",
    "    result = 1-np.log(array)/math.log(50000)\n",
    "    result[result<0]=0\n",
    "    return result\n",
    "def index_to_hotshot_encoding(index_encoded_peptide):\n",
    "    result = np.zeros((len(index_encoded_peptide),21))\n",
    "    for position, amino_acid in enumerate(index_encoded_peptide):\n",
    "        result[position][amino_acid]= 1\n",
    "    return result.flatten()\n",
    "def amino_acid_hotshot_encoding(s):\n",
    "    return common_amino_acids.hotshot_encoding([s],len(s)).flatten().astype(int)\n",
    "\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",
    "\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)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "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(index_to_hotshot_encoding,1,np.apply_along_axis(first_and_last_four,1,X))\n",
    "X_33 = np.apply_along_axis(index_to_hotshot_encoding,1,np.apply_along_axis(first_and_last_three,1,X))\n",
    "X_22 = np.apply_along_axis(index_to_hotshot_encoding,1,np.apply_along_axis(first_and_last_two,1,X))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 0 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 12s - loss: 0.0430    \n",
      "epoch 1 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 12s - loss: 0.0280    \n",
      "epoch 2 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 12s - loss: 0.0248    \n",
      "epoch 3 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0233    \n",
      "epoch 4 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0229    \n",
      "epoch 5 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0223    \n",
      "epoch 6 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0221    \n",
      "epoch 7 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 11s - loss: 0.0219    \n",
      "epoch 8 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 10s - loss: 0.0218    \n",
      "epoch 9 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0215     \n",
      "epoch 10 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0213     \n",
      "epoch 11 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 14s - loss: 0.0214    \n",
      "epoch 12 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0211     \n",
      "epoch 13 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 12s - loss: 0.0213    \n",
      "epoch 14 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 12s - loss: 0.0214    \n",
      "epoch 15 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0212    \n",
      "epoch 16 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 12s - loss: 0.0212    \n",
      "epoch 17 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0210    \n",
      "epoch 18 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0209    \n",
      "epoch 19 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 13s - loss: 0.0212    \n",
      "epoch 20 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 12s - loss: 0.0209    \n",
      "epoch 21 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 14s - loss: 0.0211    \n",
      "epoch 22 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 10s - loss: 0.0209    \n",
      "epoch 23 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0210     \n",
      "epoch 24 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0212     \n",
      "epoch 25 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0207     \n",
      "epoch 26 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 9s - loss: 0.0208     \n",
      "epoch 27 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 10s - loss: 0.0206    \n",
      "epoch 28 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 10s - loss: 0.0209    \n",
      "epoch 29 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0209     \n",
      "epoch 30 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0209     \n",
      "epoch 31 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 32 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 33 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0208     \n",
      "epoch 34 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 35 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0205     \n",
      "epoch 36 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0206     \n",
      "epoch 37 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0209     \n",
      "epoch 38 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0209     \n",
      "epoch 39 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0208     \n",
      "epoch 40 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0206     \n",
      "epoch 41 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 42 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0211     \n",
      "epoch 43 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0210     \n",
      "epoch 44 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0206     \n",
      "epoch 45 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0208     \n",
      "epoch 46 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0209     \n",
      "epoch 47 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 48 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 49 fold 0\n",
      "Epoch 1/1\n",
      "6376/6376 [==============================] - 8s - loss: 0.0208     \n",
      "epoch 0 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0435     \n",
      "epoch 1 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0274     \n",
      "epoch 2 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0247     \n",
      "epoch 3 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0234     \n",
      "epoch 4 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0227     \n",
      "epoch 5 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0221     \n",
      "epoch 6 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0220     \n",
      "epoch 7 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0218     \n",
      "epoch 8 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0217     \n",
      "epoch 9 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0214     \n",
      "epoch 10 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0211     \n",
      "epoch 11 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0210     \n",
      "epoch 12 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0208     \n",
      "epoch 13 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 12s - loss: 0.0211    \n",
      "epoch 14 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 13s - loss: 0.0212    \n",
      "epoch 15 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 9s - loss: 0.0208     \n",
      "epoch 16 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0210    \n",
      "epoch 17 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 17s - loss: 0.0208    \n",
      "epoch 18 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0208     \n",
      "epoch 19 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 9s - loss: 0.0208     \n",
      "epoch 20 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 9s - loss: 0.0208     \n",
      "epoch 21 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0209    \n",
      "epoch 22 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 9s - loss: 0.0212     \n",
      "epoch 23 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 9s - loss: 0.0206     \n",
      "epoch 24 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0209     \n",
      "epoch 25 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0208     \n",
      "epoch 26 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 12s - loss: 0.0210    \n",
      "epoch 27 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 9s - loss: 0.0208     \n",
      "epoch 28 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 13s - loss: 0.0208    \n",
      "epoch 29 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 12s - loss: 0.0207    \n",
      "epoch 30 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0208    \n",
      "epoch 31 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 9s - loss: 0.0207     \n",
      "epoch 32 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0206     \n",
      "epoch 33 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0209     \n",
      "epoch 34 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0208     \n",
      "epoch 35 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 9s - loss: 0.0205     \n",
      "epoch 36 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0205     \n",
      "epoch 37 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 9s - loss: 0.0208     \n",
      "epoch 38 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0208    \n",
      "epoch 39 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 10s - loss: 0.0208    \n",
      "epoch 40 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 41 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 42 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0205     \n",
      "epoch 43 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 44 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0208     \n",
      "epoch 45 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 46 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0204     \n",
      "epoch 47 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 48 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0203     \n",
      "epoch 49 fold 1\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0208     \n",
      "epoch 0 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0444     \n",
      "epoch 1 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0281     \n",
      "epoch 2 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0251     \n",
      "epoch 3 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0237     \n",
      "epoch 4 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0230     \n",
      "epoch 5 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0223     \n",
      "epoch 6 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0220     \n",
      "epoch 7 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0216     \n",
      "epoch 8 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0215     \n",
      "epoch 9 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0215     \n",
      "epoch 10 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0212     \n",
      "epoch 11 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0213     \n",
      "epoch 12 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0211     \n",
      "epoch 13 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0211     \n",
      "epoch 14 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0213     \n",
      "epoch 15 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0211     \n",
      "epoch 16 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0210     \n",
      "epoch 17 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0210     \n",
      "epoch 18 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0210     \n",
      "epoch 19 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0210     \n",
      "epoch 20 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 9s - loss: 0.0209     \n",
      "epoch 21 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 22 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0210     \n",
      "epoch 23 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0208     \n",
      "epoch 24 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0206     \n",
      "epoch 25 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0208     \n",
      "epoch 26 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0209     \n",
      "epoch 27 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0211     \n",
      "epoch 28 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 29 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0208     \n",
      "epoch 30 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 31 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 32 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0206     \n",
      "epoch 33 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0208     \n",
      "epoch 34 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0208     \n",
      "epoch 35 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0206     \n",
      "epoch 36 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0208     \n",
      "epoch 37 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0206     \n",
      "epoch 38 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 16s - loss: 0.0207    \n",
      "epoch 39 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 13s - loss: 0.0209    \n",
      "epoch 40 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 16s - loss: 0.0205    \n",
      "epoch 41 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 15s - loss: 0.0207    \n",
      "epoch 42 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 11s - loss: 0.0207    \n",
      "epoch 43 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0206     \n",
      "epoch 44 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0205     \n",
      "epoch 45 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 46 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 47 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0207     \n",
      "epoch 48 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0205     \n",
      "epoch 49 fold 2\n",
      "Epoch 1/1\n",
      "6377/6377 [==============================] - 8s - loss: 0.0205     \n"
     ]
    }
   ],
   "source": [
    "folds = 3\n",
    "n_epochs = 50\n",
    "batch_size_nn = 16\n",
    "batch_size_lstm = 16\n",
    "hidden = 50\n",
    "dropout_probability = 0.25\n",
    "\n",
    "train_lr_aucs = np.zeros(folds)\n",
    "test_lr_aucs = np.zeros(folds)\n",
    "nine_train_lr_aucs = np.zeros(folds)\n",
    "nine_test_lr_aucs = np.zeros(folds)\n",
    "non_nine_train_lr_aucs = np.zeros(folds)\n",
    "non_nine_test_lr_aucs = np.zeros(folds)\n",
    "\n",
    "train_lr_22_aucs = np.zeros(folds)\n",
    "test_lr_22_aucs = np.zeros(folds)\n",
    "nine_train_lr_22_aucs = np.zeros(folds)\n",
    "nine_test_lr_22_aucs = np.zeros(folds)\n",
    "non_nine_train_lr_22_aucs = np.zeros(folds)\n",
    "non_nine_test_lr_22_aucs = np.zeros(folds)\n",
    "\n",
    "train_lr_33_aucs = np.zeros(folds)\n",
    "test_lr_33_aucs = np.zeros(folds)\n",
    "nine_train_lr_33_aucs = np.zeros(folds)\n",
    "nine_test_lr_33_aucs = np.zeros(folds)\n",
    "non_nine_train_lr_33_aucs = np.zeros(folds)\n",
    "non_nine_test_lr_33_aucs = np.zeros(folds)\n",
    "\n",
    "train_lr_44_aucs = np.zeros(folds)\n",
    "test_lr_44_aucs = np.zeros(folds)\n",
    "nine_train_lr_44_aucs = np.zeros(folds)\n",
    "nine_test_lr_44_aucs = np.zeros(folds)\n",
    "non_nine_train_lr_44_aucs = np.zeros(folds)\n",
    "non_nine_test_lr_44_aucs = np.zeros(folds)\n",
    "\n",
    "train_lr_f1= np.zeros(folds)\n",
    "test_lr_f1= np.zeros(folds)\n",
    "nine_train_lr_f1= np.zeros(folds)\n",
    "nine_test_lr_f1= np.zeros(folds)\n",
    "non_nine_train_lr_f1= np.zeros(folds)\n",
    "non_nine_test_lr_f1= np.zeros(folds)\n",
    "\n",
    "train_lr_22_f1= np.zeros(folds)\n",
    "test_lr_22_f1= np.zeros(folds)\n",
    "nine_train_lr_22_f1= np.zeros(folds)\n",
    "nine_test_lr_22_f1= np.zeros(folds)\n",
    "non_nine_train_lr_22_f1= np.zeros(folds)\n",
    "non_nine_test_lr_22_f1= np.zeros(folds)\n",
    "\n",
    "train_lr_33_f1= np.zeros(folds)\n",
    "test_lr_33_f1= np.zeros(folds)\n",
    "nine_train_lr_33_f1= np.zeros(folds)\n",
    "nine_test_lr_33_f1= np.zeros(folds)\n",
    "non_nine_train_lr_33_f1= np.zeros(folds)\n",
    "non_nine_test_lr_33_f1= np.zeros(folds)\n",
    "\n",
    "train_lr_44_f1= np.zeros(folds)\n",
    "test_lr_44_f1= np.zeros(folds)\n",
    "nine_train_lr_44_f1= np.zeros(folds)\n",
    "nine_test_lr_44_f1= np.zeros(folds)\n",
    "non_nine_train_lr_44_f1= np.zeros(folds)\n",
    "non_nine_test_lr_44_f1= np.zeros(folds)\n",
    "\n",
    "train_lstm_aucs = np.zeros((folds,n_epochs))\n",
    "test_lstm_aucs = np.zeros((folds,n_epochs))\n",
    "nine_train_lstm_aucs = np.zeros((folds,n_epochs))\n",
    "nine_test_lstm_aucs = np.zeros((folds,n_epochs))\n",
    "non_nine_train_lstm_aucs = np.zeros((folds,n_epochs))\n",
    "non_nine_test_lstm_aucs = np.zeros((folds,n_epochs))\n",
    "\n",
    "\n",
    "train_lstm_f1 = np.zeros((folds,n_epochs))\n",
    "test_lstm_f1 = np.zeros((folds,n_epochs))\n",
    "nine_train_lstm_f1 = np.zeros((folds,n_epochs))\n",
    "nine_test_lstm_f1 = np.zeros((folds,n_epochs))\n",
    "non_nine_train_lstm_f1 = np.zeros((folds,n_epochs))\n",
    "non_nine_test_lstm_f1 = np.zeros((folds,n_epochs))\n",
    "\n",
    "\n",
    "\n",
    "for i, (train_idx, test_idx) in enumerate(KFold(len(ds_h),folds, shuffle=True)):\n",
    "    \n",
    "    #indices\n",
    "    \n",
    "    nine_train_idx = np.array([i for i in train_idx if (np.count_nonzero(X[i])==9)])\n",
    "    non_nine_train_idx = np.array([i for i in train_idx if (np.count_nonzero(X[i])!=9)])\n",
    "    nine_test_idx = np.array([i for i in test_idx if (np.count_nonzero(X[i])==9)])\n",
    "    non_nine_test_idx = np.array([i for i in test_idx if (np.count_nonzero(X[i])!=9)])\n",
    "    \n",
    "    \n",
    "    # training sets\n",
    "    \n",
    "    \n",
    "    X_train = slicing(ds_h,train_idx,0)\n",
    "    y_train = label_transform(slicing(ds_h,train_idx,1))\n",
    "    \n",
    "    X_train_hotshot = np.apply_along_axis(index_to_hotshot_encoding,1, X_train )\n",
    "    \n",
    "    X_nine_train = slicing(ds_h,nine_train_idx,0)\n",
    "    y_nine_train = label_transform(slicing(ds_h,nine_train_idx,1))\n",
    "    \n",
    "    X_nine_train_hotshot = np.apply_along_axis(index_to_hotshot_encoding,1, X_nine_train )\n",
    "    \n",
    "    X_non_nine_train = slicing(ds_h,non_nine_train_idx,0)\n",
    "    y_non_nine_train = label_transform(slicing(ds_h,non_nine_train_idx,1))\n",
    "    \n",
    "    X_non_nine_train_hotshot = np.apply_along_axis(index_to_hotshot_encoding,1, X_non_nine_train )\n",
    "    \n",
    "    \n",
    "    # test sets\n",
    "    \n",
    "    X_test = slicing(ds_h,test_idx,0)\n",
    "    y_test = label_transform(slicing(ds_h,test_idx,1))\n",
    "    \n",
    "    X_test_hotshot = np.apply_along_axis(index_to_hotshot_encoding,1, X_test)\n",
    "    \n",
    "    X_nine_test = slicing(ds_h,nine_test_idx,0)\n",
    "    y_nine_test = label_transform(slicing(ds_h,nine_test_idx,1))\n",
    "    \n",
    "    X_nine_test_hotshot = np.apply_along_axis(index_to_hotshot_encoding,1, X_nine_test)\n",
    "    \n",
    "    X_non_nine_test = slicing(ds_h,non_nine_test_idx,0)\n",
    "    y_non_nine_test = label_transform(slicing(ds_h,non_nine_test_idx,1))\n",
    "    \n",
    "    X_non_nine_test_hotshot = np.apply_along_axis(index_to_hotshot_encoding,1, X_non_nine_test)\n",
    "    \n",
    "    # weights\n",
    "    \n",
    "    weights_train = slicing(ds_h,train_idx,2)\n",
    "    weights_test = slicing(ds_h,test_idx,2)\n",
    "    \n",
    "    weights_nine_train = slicing(ds_h,nine_train_idx,2)\n",
    "    weights_nine_test = slicing(ds_h,nine_test_idx,2)\n",
    "    \n",
    "    weights_non_nine_train = slicing(ds_h,non_nine_train_idx,2)\n",
    "    weights_non_nine_test = slicing(ds_h,non_nine_test_idx,2)\n",
    "    \n",
    "    # original indices\n",
    "    \n",
    "    original_indices_train = slicing(ds_h,train_idx,3)\n",
    "    original_indices_test = slicing(ds_h,test_idx,3)\n",
    "    \n",
    "    original_indices_nine_train = slicing(ds_h,nine_train_idx,3)\n",
    "    original_indices_nine_test = slicing(ds_h,nine_test_idx,3)\n",
    "    \n",
    "    original_indices_non_nine_train = slicing(ds_h,non_nine_train_idx,3)\n",
    "    original_indices_non_nine_test = slicing(ds_h,non_nine_test_idx,3)\n",
    "    \n",
    "    # real labels (to be used for AUC score calculation)\n",
    "    \n",
    "    train_real_labels = regroup_together(y_train, weights_train , original_indices_train)\n",
    "    test_real_labels = regroup_together(y_test, weights_test , original_indices_test)\n",
    "    \n",
    "    nine_train_real_labels = regroup_together(y_nine_train, weights_nine_train , original_indices_nine_train)\n",
    "    nine_test_real_labels = regroup_together(y_nine_test, weights_nine_test , original_indices_nine_test)\n",
    "    \n",
    "    non_nine_train_real_labels = regroup_together(y_non_nine_train, weights_non_nine_train , original_indices_non_nine_train)\n",
    "    non_nine_test_real_labels = regroup_together(y_non_nine_test, weights_non_nine_test , original_indices_non_nine_test)\n",
    "    \n",
    "    \n",
    "    \n",
    "    lr = LinearRegression()\n",
    "    lr_22 = LinearRegression()\n",
    "    lr_33 = LinearRegression()\n",
    "    lr_44 = LinearRegression()\n",
    "    \n",
    "    lr.fit(X_train_hotshot,y_train, sample_weight = weights_train)\n",
    "    lr_22.fit(X_22[train_idx],y[train_idx])\n",
    "    lr_33.fit(X_33[train_idx],y[train_idx])\n",
    "    lr_44.fit(X_44[train_idx],y[train_idx])\n",
    "    \n",
    "    train_pred_lr = regroup_together(lr.predict(X_train_hotshot), weights_train, original_indices_train)\n",
    "    test_pred_lr = regroup_together(lr.predict(X_test_hotshot), weights_test, original_indices_test)\n",
    "    \n",
    "    nine_train_pred_lr = regroup_together(lr.predict(X_nine_train_hotshot), weights_nine_train, original_indices_nine_train)\n",
    "    nine_test_pred_lr = regroup_together(lr.predict(X_nine_test_hotshot), weights_nine_test, original_indices_nine_test)\n",
    "    \n",
    "    non_nine_train_pred_lr = regroup_together(lr.predict(X_non_nine_train_hotshot), weights_non_nine_train, original_indices_non_nine_train)\n",
    "    non_nine_test_pred_lr = regroup_together(lr.predict(X_non_nine_test_hotshot), weights_non_nine_test, original_indices_non_nine_test)\n",
    "    \n",
    "    train_lr_aucs[i] = roc_auc_score(measured_affinity_less_than(train_real_labels,500), train_pred_lr)     \n",
    "    test_lr_aucs[i] = roc_auc_score(measured_affinity_less_than(test_real_labels,500), test_pred_lr)\n",
    "    nine_train_lr_aucs[i] = roc_auc_score(measured_affinity_less_than(nine_train_real_labels,500), nine_train_pred_lr)     \n",
    "    nine_test_lr_aucs[i] = roc_auc_score(measured_affinity_less_than(nine_test_real_labels,500), nine_test_pred_lr)\n",
    "    non_nine_train_lr_aucs[i] = roc_auc_score(measured_affinity_less_than(non_nine_train_real_labels,500), non_nine_train_pred_lr)     \n",
    "    non_nine_test_lr_aucs[i] = roc_auc_score(measured_affinity_less_than(non_nine_test_real_labels,500), non_nine_test_pred_lr)\n",
    "    \n",
    "    train_lr_22_aucs[i] = roc_auc_score(measured_affinity_less_than(y[train_idx],500), lr_22.predict(X_22[train_idx]))     \n",
    "    test_lr_22_aucs[i] = roc_auc_score(measured_affinity_less_than(y[test_idx],500), lr_22.predict(X_22[test_idx]))\n",
    "    nine_train_lr_22_aucs[i] = roc_auc_score(measured_affinity_less_than(y[nine_train_idx],500), lr_22.predict(X_22[nine_train_idx]))     \n",
    "    nine_test_lr_22_aucs[i] = roc_auc_score(measured_affinity_less_than(y[nine_test_idx],500), lr_22.predict(X_22[nine_test_idx]))\n",
    "    non_nine_train_lr_22_aucs[i] = roc_auc_score(measured_affinity_less_than(y[non_nine_train_idx],500), lr_22.predict(X_22[non_nine_train_idx]))     \n",
    "    non_nine_test_lr_22_aucs[i] = roc_auc_score(measured_affinity_less_than(y[non_nine_test_idx],500), lr_22.predict(X_22[non_nine_test_idx]))\n",
    "    \n",
    "    train_lr_33_aucs[i]=  roc_auc_score(measured_affinity_less_than(y[train_idx],500), lr_33.predict(X_33[train_idx]))     \n",
    "    test_lr_33_aucs[i] = roc_auc_score(measured_affinity_less_than(y[test_idx],500), lr_33.predict(X_33[test_idx]))\n",
    "    nine_train_lr_33_aucs[i]=  roc_auc_score(measured_affinity_less_than(y[nine_train_idx],500), lr_33.predict(X_33[nine_train_idx]))     \n",
    "    nine_test_lr_33_aucs[i] = roc_auc_score(measured_affinity_less_than(y[nine_test_idx],500), lr_33.predict(X_33[nine_test_idx]))\n",
    "    non_nine_train_lr_33_aucs[i]=  roc_auc_score(measured_affinity_less_than(y[non_nine_train_idx],500), lr_33.predict(X_33[non_nine_train_idx]))     \n",
    "    non_nine_test_lr_33_aucs[i] = roc_auc_score(measured_affinity_less_than(y[non_nine_test_idx],500), lr_33.predict(X_33[non_nine_test_idx]))\n",
    "    \n",
    "    train_lr_44_aucs[i]= roc_auc_score(measured_affinity_less_than(y[train_idx],500), lr_44.predict(X_44[train_idx]))     \n",
    "    test_lr_44_aucs[i]= roc_auc_score(measured_affinity_less_than(y[test_idx],500), lr_44.predict(X_44[test_idx]))\n",
    "    nine_train_lr_44_aucs[i]= roc_auc_score(measured_affinity_less_than(y[nine_train_idx],500), lr_44.predict(X_44[nine_train_idx]))     \n",
    "    nine_test_lr_44_aucs[i]= roc_auc_score(measured_affinity_less_than(y[nine_test_idx],500), lr_44.predict(X_44[nine_test_idx]))\n",
    "    non_nine_train_lr_44_aucs[i]= roc_auc_score(measured_affinity_less_than(y[non_nine_train_idx],500), lr_44.predict(X_44[non_nine_train_idx]))     \n",
    "    non_nine_test_lr_44_aucs[i]= roc_auc_score(measured_affinity_less_than(y[non_nine_test_idx],500), lr_44.predict(X_44[non_nine_test_idx]))\n",
    "    \n",
    "    train_lr_f1[i] = f1_score(measured_affinity_less_than(train_real_labels,500),measured_affinity_less_than(train_pred_lr,500))\n",
    "    test_lr_f1[i] = f1_score(measured_affinity_less_than(test_real_labels,500), measured_affinity_less_than(test_pred_lr,500))\n",
    "    nine_train_lr_f1[i] = f1_score(measured_affinity_less_than(nine_train_real_labels,500),measured_affinity_less_than(nine_train_pred_lr,500))\n",
    "    nine_test_lr_f1[i] = f1_score(measured_affinity_less_than(nine_test_real_labels,500), measured_affinity_less_than(nine_test_pred_lr,500))\n",
    "    non_nine_train_lr_f1[i] = f1_score(measured_affinity_less_than(non_nine_train_real_labels,500),measured_affinity_less_than(non_nine_train_pred_lr,500))\n",
    "    non_nine_test_lr_f1[i] = f1_score(measured_affinity_less_than(non_nine_test_real_labels,500), measured_affinity_less_than(non_nine_test_pred_lr,500))\n",
    "    \n",
    "    train_lr_22_f1[i] = f1_score(measured_affinity_less_than(y[train_idx],500),measured_affinity_less_than(lr_22.predict(X_22[train_idx]),500))\n",
    "    test_lr_22_f1[i] = f1_score(measured_affinity_less_than(y[test_idx],500), measured_affinity_less_than(lr_22.predict(X_22[test_idx]),500))\n",
    "    nine_train_lr_22_f1[i] = f1_score(measured_affinity_less_than(y[nine_train_idx],500),measured_affinity_less_than(lr_22.predict(X_22[nine_train_idx]),500))\n",
    "    nine_test_lr_22_f1[i] = f1_score(measured_affinity_less_than(y[nine_test_idx],500), measured_affinity_less_than(lr_22.predict(X_22[nine_test_idx]),500))\n",
    "    non_nine_train_lr_22_f1[i] = f1_score(measured_affinity_less_than(y[non_nine_train_idx],500),measured_affinity_less_than(lr_22.predict(X_22[non_nine_train_idx]),500))\n",
    "    non_nine_test_lr_22_f1[i] = f1_score(measured_affinity_less_than(y[non_nine_test_idx],500), measured_affinity_less_than(lr_22.predict(X_22[non_nine_test_idx]),500))\n",
    "    \n",
    "    train_lr_33_f1[i] = f1_score(measured_affinity_less_than(y[train_idx],500),measured_affinity_less_than(lr_33.predict(X_33[train_idx]),500))\n",
    "    test_lr_33_f1[i] = f1_score(measured_affinity_less_than(y[test_idx],500), measured_affinity_less_than(lr_33.predict(X_33[test_idx]),500))\n",
    "    nine_train_lr_33_f1[i] = f1_score(measured_affinity_less_than(y[nine_train_idx],500),measured_affinity_less_than(lr_33.predict(X_33[nine_train_idx]),500))\n",
    "    nine_test_lr_33_f1[i] = f1_score(measured_affinity_less_than(y[nine_test_idx],500), measured_affinity_less_than(lr_33.predict(X_33[nine_test_idx]),500))\n",
    "    non_nine_train_lr_33_f1[i] = f1_score(measured_affinity_less_than(y[non_nine_train_idx],500),measured_affinity_less_than(lr_33.predict(X_33[non_nine_train_idx]),500))\n",
    "    non_nine_test_lr_33_f1[i] = f1_score(measured_affinity_less_than(y[non_nine_test_idx],500), measured_affinity_less_than(lr_33.predict(X_33[non_nine_test_idx]),500))\n",
    "    \n",
    "    train_lr_44_f1[i] = f1_score(measured_affinity_less_than(y[train_idx],500),measured_affinity_less_than(lr_44.predict(X_44[train_idx]),500))\n",
    "    test_lr_44_f1[i] = f1_score(measured_affinity_less_than(y[test_idx],500), measured_affinity_less_than(lr_44.predict(X_44[test_idx]),500))\n",
    "    nine_train_lr_44_f1[i] = f1_score(measured_affinity_less_than(y[nine_train_idx],500),measured_affinity_less_than(lr_44.predict(X_44[nine_train_idx]),500))\n",
    "    nine_test_lr_44_f1[i] = f1_score(measured_affinity_less_than(y[nine_test_idx],500), measured_affinity_less_than(lr_44.predict(X_44[nine_test_idx]),500))\n",
    "    non_nine_train_lr_44_f1[i] = f1_score(measured_affinity_less_than(y[non_nine_train_idx],500),measured_affinity_less_than(lr_44.predict(X_44[non_nine_train_idx]),500))\n",
    "    non_nine_test_lr_44_f1[i] = f1_score(measured_affinity_less_than(y[non_nine_test_idx],500), measured_affinity_less_than(lr_44.predict(X_44[non_nine_test_idx]),500))\n",
    "    \n",
    "    \n",
    "\n",
    "    \n",
    "    # normal LSTM\n",
    "    \n",
    "    sequence = Input( shape= (26, ),dtype='int32')\n",
    "    embedded = Embedding(input_dim = 21, input_length = 26, 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",
    "    \n",
    "    for epoch in range(n_epochs):\n",
    "        print(\"epoch\", epoch,\"fold\", i)\n",
    "        adam.lr.set_value(0.01*(epoch+1)**(-2))\n",
    "        lstm.fit(X[train_idx],y[train_idx], batch_size = batch_size_lstm, nb_epoch=1)\n",
    "    \n",
    "\n",
    "        train_lstm_auc = roc_auc_score(measured_affinity_less_than(y[train_idx],500),lstm.predict(X[train_idx]))\n",
    "        test_lstm_auc = roc_auc_score(measured_affinity_less_than(y[test_idx],500),lstm.predict(X[test_idx]))\n",
    "        nine_train_lstm_auc = roc_auc_score(measured_affinity_less_than(y[nine_train_idx],500),lstm.predict(X[nine_train_idx]))\n",
    "        nine_test_lstm_auc = roc_auc_score(measured_affinity_less_than(y[nine_test_idx],500),lstm.predict(X[nine_test_idx]))\n",
    "        non_nine_train_lstm_auc = roc_auc_score(measured_affinity_less_than(y[non_nine_train_idx],500),lstm.predict(X[non_nine_train_idx]))\n",
    "        non_nine_test_lstm_auc = roc_auc_score(measured_affinity_less_than(y[non_nine_test_idx],500),lstm.predict(X[non_nine_test_idx]))\n",
    "        \n",
    "        train_lstm_aucs[i][epoch]=train_lstm_auc\n",
    "        test_lstm_aucs[i][epoch]=test_lstm_auc\n",
    "        nine_train_lstm_aucs[i][epoch]=nine_train_lstm_auc\n",
    "        nine_test_lstm_aucs[i][epoch]=nine_test_lstm_auc\n",
    "        non_nine_train_lstm_aucs[i][epoch]=non_nine_train_lstm_auc\n",
    "        non_nine_test_lstm_aucs[i][epoch]=non_nine_test_lstm_auc\n",
    "        \n",
    "        train_lstm_f1[i][epoch] = f1_score(measured_affinity_less_than(y[train_idx],500),measured_affinity_less_than(lstm.predict(X[train_idx]),500))\n",
    "        test_lstm_f1[i][epoch] = f1_score(measured_affinity_less_than(y[test_idx],500), measured_affinity_less_than(lstm.predict(X[test_idx]),500))\n",
    "        nine_train_lstm_f1[i][epoch] = f1_score(measured_affinity_less_than(y[nine_train_idx],500),measured_affinity_less_than(lstm.predict(X[nine_train_idx]),500))\n",
    "        nine_test_lstm_f1[i][epoch] = f1_score(measured_affinity_less_than(y[nine_test_idx],500), measured_affinity_less_than(lstm.predict(X[nine_test_idx]),500))\n",
    "        non_nine_train_lstm_f1[i][epoch] = f1_score(measured_affinity_less_than(y[non_nine_train_idx],500),measured_affinity_less_than(lstm.predict(X[non_nine_train_idx]),500))\n",
    "        non_nine_test_lstm_f1[i][epoch] = f1_score(measured_affinity_less_than(y[non_nine_test_idx],500), measured_affinity_less_than(lstm.predict(X[non_nine_test_idx]),500))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "test_lr_44_aucs_mean = np.mean(test_lr_44_aucs, axis=0)\n",
    "test_lr_33_aucs_mean = np.mean(test_lr_33_aucs, axis=0)\n",
    "test_lr_22_aucs_mean = np.mean(test_lr_22_aucs, axis=0)\n",
    "test_lr_aucs_mean = np.mean(test_lr_aucs, axis=0)\n",
    "test_lstm_aucs_mean = np.mean(test_lstm_aucs, axis=0)\n",
    "\n",
    "test_lr_44_f1_mean = np.mean(test_lr_44_f1, axis=0)\n",
    "test_lr_33_f1_mean = np.mean(test_lr_33_f1, axis=0)\n",
    "test_lr_22_f1_mean = np.mean(test_lr_22_f1, axis=0)\n",
    "test_lr_f1_mean = np.mean(test_lr_f1, axis=0)\n",
    "test_lstm_f1_mean = np.mean(test_lstm_f1, axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 0.94007066,  0.94007066,  0.94007066,  0.94007066,  0.94007066,\n",
       "        0.94007066,  0.94007066,  0.94007066,  0.94007066,  0.94007066,\n",
       "        0.94007066,  0.94007066,  0.94007066,  0.94007066,  0.94007066,\n",
       "        0.94007066,  0.94007066,  0.94007066,  0.94007066,  0.94007066,\n",
       "        0.94007066,  0.94007066,  0.94007066,  0.94007066,  0.94007066,\n",
       "        0.94007066,  0.94007066,  0.94007066,  0.94007066,  0.94007066,\n",
       "        0.94007066,  0.94007066,  0.94007066,  0.94007066,  0.94007066,\n",
       "        0.94007066,  0.94007066,  0.94007066,  0.94007066,  0.94007066,\n",
       "        0.94007066,  0.94007066,  0.94007066,  0.94007066,  0.94007066,\n",
       "        0.94007066,  0.94007066,  0.94007066,  0.94007066,  0.94007066])"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.full(n_epochs, test_lr_44_aucs_mean)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.943143747902 0.940070657535 0.9301802046 0.897164945373\n"
     ]
    }
   ],
   "source": [
    "print(test_lr_aucs_mean, test_lr_44_aucs_mean, test_lr_33_aucs_mean, test_lr_22_aucs_mean)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.820123228547 0.816675958356 0.792619977087 0.738639177717\n"
     ]
    }
   ],
   "source": [
    "print(test_lr_f1_mean, test_lr_44_f1_mean, test_lr_33_f1_mean, test_lr_22_f1_mean)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x139ded7f0>"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAhEAAAFoCAYAAADpW6mGAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nIiISFCURIiIiEpT/BzZJ2hXPlI2LAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1346063c8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot( np.arange(0,50,1), np.full(n_epochs, test_lr_aucs_mean),color='b',marker='*', linestyle='-', label =\"test linear\")\n",
    "plt.plot( np.arange(0,50,1), np.full(n_epochs, test_lr_44_aucs_mean),color='g',marker='*', linestyle='-', label =\"test linear 44\")\n",
    "plt.plot( np.arange(0,50,1), np.full(n_epochs,test_lr_33_aucs_mean),color='orange',marker='*', linestyle='-', label =\"test linear 33\")\n",
    "plt.plot( np.arange(0,50,1), np.full(n_epochs,test_lr_22_aucs_mean),color='r',marker='*', linestyle='-', label =\"test linear 22\")\n",
    "plt.plot( np.arange(0,50,1), test_lstm_aucs_mean,color='c',marker='*', linestyle='-', label =\"test LSTM\")\n",
    "\n",
    "plt.legend(loc=4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "nine_test_lr_44_aucs_mean = np.mean(nine_test_lr_44_aucs, axis=0)\n",
    "nine_test_lr_33_aucs_mean = np.mean(nine_test_lr_33_aucs, axis=0)\n",
    "nine_test_lr_22_aucs_mean = np.mean(nine_test_lr_22_aucs, axis=0)\n",
    "nine_test_lr_aucs_mean = np.mean(nine_test_lr_aucs, axis=0)\n",
    "nine_test_lstm_aucs_mean = np.mean(nine_test_lstm_aucs, axis=0)\n",
    "\n",
    "nine_test_lr_44_f1_mean = np.mean(nine_test_lr_44_f1, axis=0)\n",
    "nine_test_lr_33_f1_mean = np.mean(nine_test_lr_33_f1, axis=0)\n",
    "nine_test_lr_22_f1_mean = np.mean(nine_test_lr_22_f1, axis=0)\n",
    "nine_test_lr_f1_mean = np.mean(nine_test_lr_f1, axis=0)\n",
    "nine_test_lstm_f1_mean = np.mean(nine_test_lstm_f1, axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1365d3710>"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAhEAAAFoCAYAAADpW6mGAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nUSD/D/OUNaFCB6AsAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1367ed828>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot( np.arange(0,50,1), np.full(n_epochs, nine_test_lr_aucs_mean),color='b',marker='*', linestyle='-', label =\"test linear on 9 mers\")\n",
    "plt.plot( np.arange(0,50,1), np.full(n_epochs, nine_test_lr_44_aucs_mean),color='g',marker='*', linestyle='-', label =\"test linear 44 on 9 mers\")\n",
    "plt.plot( np.arange(0,50,1), np.full(n_epochs,nine_test_lr_33_aucs_mean),color='orange',marker='*', linestyle='-', label =\"test linear 33 on 9 mers\")\n",
    "plt.plot( np.arange(0,50,1), np.full(n_epochs,nine_test_lr_22_aucs_mean),color='r',marker='*', linestyle='-', label =\"test linear 22 on 9 mers\")\n",
    "plt.plot( np.arange(0,50,1), nine_test_lstm_aucs_mean,color='c',marker='*', linestyle='-', label =\"test LSTM on 9 mers\")\n",
    "\n",
    "plt.legend(loc=4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "non_nine_test_lr_44_aucs_mean = np.mean(non_nine_test_lr_44_aucs, axis=0)\n",
    "non_nine_test_lr_33_aucs_mean = np.mean(non_nine_test_lr_33_aucs, axis=0)\n",
    "non_nine_test_lr_22_aucs_mean = np.mean(non_nine_test_lr_22_aucs, axis=0)\n",
    "non_nine_test_lr_aucs_mean = np.mean(non_nine_test_lr_aucs, axis=0)\n",
    "non_nine_test_lstm_aucs_mean = np.mean(non_nine_test_lstm_aucs, axis=0)\n",
    "\n",
    "non_nine_test_lr_44_f1_mean = np.mean(non_nine_test_lr_44_f1, axis=0)\n",
    "non_nine_test_lr_33_f1_mean = np.mean(non_nine_test_lr_33_f1, axis=0)\n",
    "non_nine_test_lr_22_f1_mean = np.mean(non_nine_test_lr_22_f1, axis=0)\n",
    "non_nine_test_lr_f1_mean = np.mean(non_nine_test_lr_f1, axis=0)\n",
    "non_nine_test_lstm_f1_mean = np.mean(non_nine_test_lstm_f1, axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x12ddfbba8>"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAhEAAAFoCAYAAADpW6mGAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x126fe9898>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot( np.arange(0,50,1), np.full(n_epochs, non_nine_test_lr_aucs_mean),color='b',marker='*', linestyle='-', label =\"test linear on non 9 mers\")\n",
    "plt.plot( np.arange(0,50,1), np.full(n_epochs, non_nine_test_lr_44_aucs_mean),color='g',marker='*', linestyle='-', label =\"test linear 44 on non 9 mers\")\n",
    "plt.plot( np.arange(0,50,1), np.full(n_epochs,non_nine_test_lr_33_aucs_mean),color='orange',marker='*', linestyle='-', label =\"test linear 33 on non 9 mers\")\n",
    "plt.plot( np.arange(0,50,1), np.full(n_epochs,non_nine_test_lr_22_aucs_mean),color='r',marker='*', linestyle='-', label =\"test linear 22 on non 9 mers\")\n",
    "plt.plot( np.arange(0,50,1), non_nine_test_lstm_aucs_mean,color='c',marker='*', linestyle='-', label =\"test LSTM on non 9 mers\")\n",
    "\n",
    "plt.legend(loc=4)"
   ]
  },
  {
   "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": 0
}