semi-supervised-trick.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# semi-supervides trick"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Populating the interactive namespace from numpy and matplotlib\n"
     ]
    }
   ],
   "source": [
    "%pylab inline\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "import scipy\n",
    "import time\n",
    "import random\n",
    "import cPickle\n",
    "\n",
    "from sklearn.externals import joblib\n",
    "from sklearn.cross_validation import KFold\n",
    "\n",
    "import set_up\n",
    "\n",
    "import sys\n",
    "sys.path.append('../git/xgboost/wrapper/')\n",
    "import xgboost as xgb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10868, 203) (10868,)\n",
      "(10873, 203)\n"
     ]
    }
   ],
   "source": [
    "X_base_tr, X_base_te = joblib.load(set_up.feats_folder_path + 'X_basepack')\n",
    "X_tr, X_te = cPickle.load(open(set_up.feats_folder_path + '4gr_pack_dim100.pickled', 'rb'))\n",
    "\n",
    "Xtrain = np.hstack((X_base_tr, X_tr))\n",
    "Xtest = np.hstack((X_base_te, X_te))\n",
    "\n",
    "ytrain = np.array(pd.read_csv(set_up.train_labels_path)['Class']) - 1\n",
    "\n",
    "print shape(Xtrain), shape(ytrain)\n",
    "print shape(Xtest)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "dbest = pd.read_csv(set_up.submissions_dir_path + 'release0.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "Y = np.array(ytrain)\n",
    "test_preds = np.array(dbest.ix[:, 1:])\n",
    "\n",
    "x_train = np.array(Xtrain)\n",
    "x_test = np.array(Xtest[:, :])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "rvs = [scipy.stats.rv_discrete(values = (range(9),test_preds[j,:])) for j in range(test_preds.shape[0])]\n",
    " \n",
    "skf = KFold(test_preds.shape[0], n_folds=10,random_state = 11)\n",
    " \n",
    "loss = 0\n",
    "test_pr1 = test_preds.copy()\n",
    " \n",
    "t0 = time.time()\n",
    "    \n",
    "kok = 0\n",
    "for train_index, test_index in skf:\n",
    " \n",
    "    datas_train = np.concatenate([x_train,x_test[train_index,:]])\n",
    "    datas_test  = x_test[test_index,:]\n",
    " \n",
    "    pr_vals = []\n",
    "    for i in range(20):\n",
    "        Y_test_s = [r.rvs() for r in rvs]\n",
    " \n",
    "        y_train  = np.array(Y.tolist()  + np.array(Y_test_s)[train_index].tolist())\n",
    "    \n",
    "        indbigtrain = np.arange(len(datas_train))\n",
    "        indtrain = np.arange(len(train_index))\n",
    "        newindtrain = random.sample(indbigtrain, int(len(indbigtrain) * 1.0))\n",
    "\n",
    "        for j in range(int(len(indtrain) * 7.0)):\n",
    "            newindtrain.append(random.choice(indtrain))\n",
    "\n",
    "        dtrain = xgb.DMatrix(datas_train[newindtrain], label=y_train[newindtrain])\n",
    "        dval = xgb.DMatrix(datas_test)\n",
    "\n",
    "        param1 = {'num_round': 200,\n",
    "                     'seed' : i*13141 + 123,\n",
    "                     'max_depth':3,\n",
    "                     'gamma': 0.0,\n",
    "                     'eta':0.22,\n",
    "                     'silent':1, \n",
    "                     'objective':'multi:softprob',\n",
    "                     'num_class' : 9,\n",
    "                     'subsample' : 1,\n",
    "                     'colsample_bytree' : 0.3,\n",
    "                     'nthread' : 16}        \n",
    "        \n",
    "        watchlist = []\n",
    " \n",
    "        bst = xgb.train(\n",
    "            param1, dtrain, param1['num_round'], watchlist)\n",
    " \n",
    "        pr_val = bst.predict(dval)\n",
    " \n",
    "        pr_vals.append(pr_val)\n",
    "        print \" - %d\" % i, (time.time() - t0) / 60\n",
    " \n",
    "    pr_val = np.mean(pr_vals,axis = 0)\n",
    " \n",
    "    test_pr1[test_index,:] = pr_val\n",
    "     \n",
    "    print ' f', kok\n",
    "    \n",
    "    kok +=1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "dfinal = pd.read_csv(set_up.test_sample_path)\n",
    "\n",
    "dfinal.ix[:, 1:] = test_pr1\n",
    "dfinal.to_csv(set_up.submissions_dir_path + 'release1.csv', index = False)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}