embeddings_classification.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import h5py\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from tensorflow.python.keras.models import Sequential\n",
    "from statsmodels.stats.proportion import proportion_confint"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Loading the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[(72298,), (72298, 150, 150, 1)]\n",
      "[(19611,), (19611, 150, 150, 1)]\n"
     ]
    }
   ],
   "source": [
    "with h5py.File('/data/joint_scoring/labels_train.h5', 'r') as hf:\n",
    "    labels_train = hf['labels_train'][:]\n",
    "    \n",
    "with h5py.File('/data/joint_scoring/img_train.h5', 'r') as hf:\n",
    "    img_train = hf['img_train'][:]\n",
    "\n",
    "with h5py.File('/data/joint_scoring/labels_test.h5', 'r') as hf:\n",
    "    labels_test = hf['labels_test'][:]\n",
    "    \n",
    "with h5py.File('/data/joint_scoring/img_test.h5', 'r') as hf:\n",
    "    img_test = hf['img_test'][:]\n",
    "\n",
    "print([labels_train.shape, img_train.shape])\n",
    "print([labels_test.shape, img_test.shape])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.python.keras.models import load_model\n",
    "\n",
    "loaded_model = load_model(\"/data/joint_scoring/models/20171108_195245.h5\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-4-e09909f2dd4a>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mpred\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mloaded_model\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpredict_classes\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mimg_test\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m/usr/local/lib/python3.5/dist-packages/tensorflow/python/keras/_impl/keras/models.py\u001b[0m in \u001b[0;36mpredict_classes\u001b[0;34m(self, x, batch_size, verbose)\u001b[0m\n\u001b[1;32m    984\u001b[0m         \u001b[0mA\u001b[0m \u001b[0mnumpy\u001b[0m \u001b[0marray\u001b[0m \u001b[0mof\u001b[0m \u001b[0;32mclass\u001b[0m \u001b[0mpredictions\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    985\u001b[0m     \"\"\"\n\u001b[0;32m--> 986\u001b[0;31m     \u001b[0mproba\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpredict\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbatch_size\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mbatch_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mverbose\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mverbose\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    987\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mproba\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mshape\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m>\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    988\u001b[0m       \u001b[0;32mreturn\u001b[0m \u001b[0mproba\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0margmax\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0maxis\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.5/dist-packages/tensorflow/python/keras/_impl/keras/models.py\u001b[0m in \u001b[0;36mpredict\u001b[0;34m(self, x, batch_size, verbose)\u001b[0m\n\u001b[1;32m    885\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbuilt\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    886\u001b[0m       \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbuild\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 887\u001b[0;31m     \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpredict\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbatch_size\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mbatch_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mverbose\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mverbose\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    888\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    889\u001b[0m   \u001b[0;32mdef\u001b[0m \u001b[0mpredict_on_batch\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.5/dist-packages/tensorflow/python/keras/_impl/keras/engine/training.py\u001b[0m in \u001b[0;36mpredict\u001b[0;34m(self, x, batch_size, verbose, steps)\u001b[0m\n\u001b[1;32m   1737\u001b[0m     \u001b[0mf\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpredict_function\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1738\u001b[0m     return self._predict_loop(\n\u001b[0;32m-> 1739\u001b[0;31m         f, ins, batch_size=batch_size, verbose=verbose, steps=steps)\n\u001b[0m\u001b[1;32m   1740\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1741\u001b[0m   \u001b[0;32mdef\u001b[0m \u001b[0mtrain_on_batch\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msample_weight\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mclass_weight\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.5/dist-packages/tensorflow/python/keras/_impl/keras/engine/training.py\u001b[0m in \u001b[0;36m_predict_loop\u001b[0;34m(self, f, ins, batch_size, verbose, steps)\u001b[0m\n\u001b[1;32m   1278\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1279\u001b[0m           \u001b[0mins_batch\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_slice_arrays\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mins\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbatch_ids\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1280\u001b[0;31m         \u001b[0mbatch_outs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mf\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mins_batch\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1281\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mbatch_outs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlist\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1282\u001b[0m           \u001b[0mbatch_outs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mbatch_outs\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.5/dist-packages/tensorflow/python/keras/_impl/keras/backend.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, inputs)\u001b[0m\n\u001b[1;32m   2475\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0moutputs\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mupdates_op\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2476\u001b[0m         \u001b[0mfeed_dict\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mfeed_dict\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2477\u001b[0;31m         **self.session_kwargs)\n\u001b[0m\u001b[1;32m   2478\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0mupdated\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0moutputs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2479\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.5/dist-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36mrun\u001b[0;34m(self, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m    887\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    888\u001b[0m       result = self._run(None, fetches, feed_dict, options_ptr,\n\u001b[0;32m--> 889\u001b[0;31m                          run_metadata_ptr)\n\u001b[0m\u001b[1;32m    890\u001b[0m       \u001b[0;32mif\u001b[0m \u001b[0mrun_metadata\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    891\u001b[0m         \u001b[0mproto_data\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtf_session\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mTF_GetBuffer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrun_metadata_ptr\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.5/dist-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m_run\u001b[0;34m(self, handle, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m   1118\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mfinal_fetches\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0mfinal_targets\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mhandle\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0mfeed_dict_tensor\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1119\u001b[0m       results = self._do_run(handle, final_targets, final_fetches,\n\u001b[0;32m-> 1120\u001b[0;31m                              feed_dict_tensor, options, run_metadata)\n\u001b[0m\u001b[1;32m   1121\u001b[0m     \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1122\u001b[0m       \u001b[0mresults\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.5/dist-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m_do_run\u001b[0;34m(self, handle, target_list, fetch_list, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m   1315\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mhandle\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1316\u001b[0m       return self._do_call(_run_fn, self._session, feeds, fetches, targets,\n\u001b[0;32m-> 1317\u001b[0;31m                            options, run_metadata)\n\u001b[0m\u001b[1;32m   1318\u001b[0m     \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1319\u001b[0m       \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_do_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0m_prun_fn\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_session\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhandle\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeeds\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfetches\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.5/dist-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m_do_call\u001b[0;34m(self, fn, *args)\u001b[0m\n\u001b[1;32m   1321\u001b[0m   \u001b[0;32mdef\u001b[0m \u001b[0m_do_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1322\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1323\u001b[0;31m       \u001b[0;32mreturn\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1324\u001b[0m     \u001b[0;32mexcept\u001b[0m \u001b[0merrors\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mOpError\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1325\u001b[0m       \u001b[0mmessage\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcompat\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mas_text\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0me\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmessage\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.5/dist-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m_run_fn\u001b[0;34m(session, feed_dict, fetch_list, target_list, options, run_metadata)\u001b[0m\n\u001b[1;32m   1300\u001b[0m           return tf_session.TF_Run(session, options,\n\u001b[1;32m   1301\u001b[0m                                    \u001b[0mfeed_dict\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfetch_list\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtarget_list\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1302\u001b[0;31m                                    status, run_metadata)\n\u001b[0m\u001b[1;32m   1303\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1304\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_prun_fn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msession\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhandle\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeed_dict\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfetch_list\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "pred = loaded_model.predict_classes(img_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "diff = pred - labels_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 0.,  0.,  3., ...,  1.,  2.,  0.])"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "diff"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Build truncated model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "config = loaded_model.get_config()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "config = config[0:48]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Sequential.from_config(config)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "weights = loaded_model.get_weights()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.set_weights(weights)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Get embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "img_combined = np.concatenate((img_train, img_test))\n",
    "labels_combined = np.concatenate((labels_train, labels_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "embeddings = model.predict(img_combined)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.decomposition import PCA\n",
    "\n",
    "x_50d = PCA(n_components=50).fit_transform(embeddings)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-13-94c1ddca9ea5>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0msklearn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmanifold\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mTSNE\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      2\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 3\u001b[0;31m \u001b[0mx_2d\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mTSNE\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mn_components\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m2\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrandom_state\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfit_transform\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx_50d\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m/usr/local/lib/python3.5/dist-packages/sklearn/manifold/t_sne.py\u001b[0m in \u001b[0;36mfit_transform\u001b[0;34m(self, X, y)\u001b[0m\n\u001b[1;32m    856\u001b[0m             \u001b[0mEmbedding\u001b[0m \u001b[0mof\u001b[0m \u001b[0mthe\u001b[0m \u001b[0mtraining\u001b[0m \u001b[0mdata\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mlow\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0mdimensional\u001b[0m \u001b[0mspace\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    857\u001b[0m         \"\"\"\n\u001b[0;32m--> 858\u001b[0;31m         \u001b[0membedding\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_fit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    859\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0membedding_\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0membedding\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    860\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0membedding_\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.5/dist-packages/sklearn/manifold/t_sne.py\u001b[0m in \u001b[0;36m_fit\u001b[0;34m(self, X, skip_num_points)\u001b[0m\n\u001b[1;32m    768\u001b[0m                           \u001b[0mX_embedded\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mX_embedded\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    769\u001b[0m                           \u001b[0mneighbors\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mneighbors_nn\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 770\u001b[0;31m                           skip_num_points=skip_num_points)\n\u001b[0m\u001b[1;32m    771\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    772\u001b[0m     \u001b[0;34m@\u001b[0m\u001b[0mproperty\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.5/dist-packages/sklearn/manifold/t_sne.py\u001b[0m in \u001b[0;36m_tsne\u001b[0;34m(self, P, degrees_of_freedom, n_samples, X_embedded, neighbors, skip_num_points)\u001b[0m\n\u001b[1;32m    825\u001b[0m             \u001b[0mopt_args\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'n_iter_without_progress'\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mn_iter_without_progress\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    826\u001b[0m             params, kl_divergence, it = _gradient_descent(obj_func, params,\n\u001b[0;32m--> 827\u001b[0;31m                                                           **opt_args)\n\u001b[0m\u001b[1;32m    828\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    829\u001b[0m         \u001b[0;31m# Save the final number of iterations\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.5/dist-packages/sklearn/manifold/t_sne.py\u001b[0m in \u001b[0;36m_gradient_descent\u001b[0;34m(objective, p0, it, n_iter, n_iter_check, n_iter_without_progress, momentum, learning_rate, min_gain, min_grad_norm, verbose, args, kwargs)\u001b[0m\n\u001b[1;32m    337\u001b[0m     \u001b[0mtic\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtime\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    338\u001b[0m     \u001b[0;32mfor\u001b[0m \u001b[0mi\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mrange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mit\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mn_iter\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 339\u001b[0;31m         \u001b[0merror\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrad\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mobjective\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mp\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    340\u001b[0m         \u001b[0mgrad_norm\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlinalg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnorm\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgrad\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    341\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.5/dist-packages/sklearn/manifold/t_sne.py\u001b[0m in \u001b[0;36m_kl_divergence_bh\u001b[0;34m(params, P, degrees_of_freedom, n_samples, n_components, angle, skip_num_points, verbose)\u001b[0m\n\u001b[1;32m    245\u001b[0m     error = _barnes_hut_tsne.gradient(val_P, X_embedded, neighbors, indptr,\n\u001b[1;32m    246\u001b[0m                                       \u001b[0mgrad\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mangle\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mn_components\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mverbose\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 247\u001b[0;31m                                       dof=degrees_of_freedom)\n\u001b[0m\u001b[1;32m    248\u001b[0m     \u001b[0mc\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m2.0\u001b[0m \u001b[0;34m*\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mdegrees_of_freedom\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0;36m1.0\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m/\u001b[0m \u001b[0mdegrees_of_freedom\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    249\u001b[0m     \u001b[0mgrad\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mgrad\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mravel\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "from sklearn.manifold import TSNE\n",
    "\n",
    "x_2d = TSNE(n_components=2, random_state=1).fit_transform(x_50d)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2D"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "from matplotlib import pyplot as plt\n",
    "\n",
    "plt.figure(figsize=(25,25))\n",
    "plt.scatter(x_2d[:, 0], x_2d[:, 1], c=y, cmap=plt.cm.get_cmap(\"viridis\", 6))\n",
    "plt.colorbar()\n",
    "plt.savefig('/data/joint_scoring/images/tsne_classification.png', bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from matplotlib import pyplot as plt\n",
    "\n",
    "plt.figure(figsize=(25,25))\n",
    "plt.scatter(x_2d[:, 0], x_2d[:, 1], c=diff, cmap=plt.cm.get_cmap(\"RdBu\", 10))\n",
    "plt.colorbar()\n",
    "plt.savefig('/data/joint_scoring/images/tsne_diference_classification.png', bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## KNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.model_selection import cross_val_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "weights = {0: 0.2468182438891165, 1: 0.61240428271328862, 2: 7.3563288563288562, 3: 15.980990274093722, 4: 27.700383141762451, 5: 12.358632478632479}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "def weight(x):\n",
    "    return np.vectorize(weights.get)(x.astype(\"int\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "embeddings_train = embeddings[:72298]\n",
    "embeddings_test = embeddings[72298:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "11\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "21\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "31\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "41\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "51\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "61\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "71\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "81\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "91\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "101\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "111\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "121\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "131\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "141\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "151\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "161\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "171\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "181\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "191\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n",
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    }
   ],
   "source": [
    "cv_scores = []\n",
    "\n",
    "kk = range(1,200,10)\n",
    "\n",
    "for k in kk:\n",
    "    print(k)\n",
    "    knn = KNeighborsClassifier(n_neighbors=k, weights=weight)\n",
    "    scores = cross_val_score(knn, embeddings_train, labels_train, cv=2, scoring='accuracy')\n",
    "    cv_scores.append(scores.mean())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.64444383555479845,\n",
       " 0.71064206686174614,\n",
       " 0.71851225123210238,\n",
       " 0.72047635806751742,\n",
       " 0.72238512685898193,\n",
       " 0.72264794602671567,\n",
       " 0.72185953367368905,\n",
       " 0.72000608685577405,\n",
       " 0.71903786278356807,\n",
       " 0.71754405623966999,\n",
       " 0.71688013439753062,\n",
       " 0.71588425852163629,\n",
       " 0.71484688963506571,\n",
       " 0.71461176742119492,\n",
       " 0.71412766877709299,\n",
       " 0.7143074766699864,\n",
       " 0.71422447151720347,\n",
       " 0.71429363010619773,\n",
       " 0.71440429471524092,\n",
       " 0.71409999707671745]"
      ]
     },
     "execution_count": 86,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cv_scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The optimal number of neighbors is 51\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZEAAAEKCAYAAADTgGjXAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7faec65bcbe0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# determining best k\n",
    "from matplotlib import pyplot as plt\n",
    "optimal_k = kk[cv_scores.index(max(cv_scores))]\n",
    "print(\"The optimal number of neighbors is %d\" % optimal_k)\n",
    "\n",
    "# plot misclassification error vs k\n",
    "plt.close()\n",
    "plt.plot(kk, cv_scores)\n",
    "plt.xlabel('Number of Neighbors K')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/sklearn/utils/extmath.py:439: RuntimeWarning: invalid value encountered in greater\n",
      "  mostfrequent = np.where(counts > oldcounts, score, oldmostfreq)\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import cross_val_predict\n",
    "knn = KNeighborsClassifier(n_neighbors=51, weights=weight)\n",
    "knn = knn.fit(X=embeddings_train, y=labels_train)\n",
    "pred = knn.predict(X=embeddings_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.70858191831115191"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.metrics import accuracy_score\n",
    "accuracy_score(labels_test, pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: statsmodels in /usr/local/lib/python3.5/dist-packages\n",
      "Requirement already satisfied: patsy in /usr/local/lib/python3.5/dist-packages (from statsmodels)\n",
      "Requirement already satisfied: scipy in /usr/local/lib/python3.5/dist-packages (from statsmodels)\n",
      "Requirement already satisfied: pandas in /usr/local/lib/python3.5/dist-packages (from statsmodels)\n",
      "Requirement already satisfied: six in /usr/local/lib/python3.5/dist-packages (from patsy->statsmodels)\n",
      "Requirement already satisfied: numpy in /usr/local/lib/python3.5/dist-packages (from patsy->statsmodels)\n",
      "Requirement already satisfied: python-dateutil>=2 in /usr/local/lib/python3.5/dist-packages (from pandas->statsmodels)\n",
      "Requirement already satisfied: pytz>=2011k in /usr/local/lib/python3.5/dist-packages (from pandas->statsmodels)\n"
     ]
    }
   ],
   "source": [
    "!pip3 install statsmodels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.70218164251235082, 0.714900494866896)"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from statsmodels.stats.proportion import proportion_confint\n",
    "proportion_confint(sum(labels_test == pred), len(labels_test), alpha=0.05, method='wilson')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion matrix, without normalization\n",
      "[[12160  1085     7    10     4    20]\n",
      " [ 3630  1594    67    49     9    20]\n",
      " [   92   221    29    56     8    17]\n",
      " [   20    85    12    38    10    12]\n",
      " [   14    39     5    17    13    10]\n",
      " [   60    90     2    31    13    62]]\n",
      "Normalized confusion matrix\n",
      "[[  9.15249134e-01   8.16649104e-02   5.26870390e-04   7.52671986e-04\n",
      "    3.01068794e-04   1.50534397e-03]\n",
      " [  6.76103557e-01   2.96889551e-01   1.24790464e-02   9.12646675e-03\n",
      "    1.67628981e-03   3.72508847e-03]\n",
      " [  2.17494090e-01   5.22458629e-01   6.85579196e-02   1.32387707e-01\n",
      "    1.89125296e-02   4.01891253e-02]\n",
      " [  1.12994350e-01   4.80225989e-01   6.77966102e-02   2.14689266e-01\n",
      "    5.64971751e-02   6.77966102e-02]\n",
      " [  1.42857143e-01   3.97959184e-01   5.10204082e-02   1.73469388e-01\n",
      "    1.32653061e-01   1.02040816e-01]\n",
      " [  2.32558140e-01   3.48837209e-01   7.75193798e-03   1.20155039e-01\n",
      "    5.03875969e-02   2.40310078e-01]]\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAVgAAAEmCAYAAAAnRIjxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEAR14n/OJ/xrH8mNzAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f46cf788a58>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUgAAAEmCAYAAAAA6gkZAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nX0xgAIlIEgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f46a1e23160>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from matplotlib import pyplot as plt\n",
    "from sklearn.metrics import confusion_matrix\n",
    "import itertools\n",
    "cnf_matrix = confusion_matrix(labels_test, pred)\n",
    "def plot_confusion_matrix(cm, classes,\n",
    "                          normalize=False,\n",
    "                          title='Confusion matrix',\n",
    "                          cmap=plt.cm.Blues):\n",
    "    \"\"\"\n",
    "    This function prints and plots the confusion matrix.\n",
    "    Normalization can be applied by setting `normalize=True`.\n",
    "    \"\"\"\n",
    "    if normalize:\n",
    "        cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n",
    "        print(\"Normalized confusion matrix\")\n",
    "    else:\n",
    "        print('Confusion matrix, without normalization')\n",
    "\n",
    "    print(cm)\n",
    "\n",
    "    plt.imshow(cm, interpolation='nearest', cmap=cmap)\n",
    "    plt.title(title)\n",
    "    plt.colorbar()\n",
    "    tick_marks = np.arange(len(classes))\n",
    "    plt.xticks(tick_marks, classes, rotation=45)\n",
    "    plt.yticks(tick_marks, classes)\n",
    "\n",
    "    fmt = '.2f' if normalize else 'd'\n",
    "    thresh = cm.max() / 2.\n",
    "    for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):\n",
    "        plt.text(j, i, format(cm[i, j], fmt),\n",
    "                 horizontalalignment=\"center\",\n",
    "                 color=\"white\" if cm[i, j] > thresh else \"black\")\n",
    "\n",
    "    plt.tight_layout()\n",
    "    plt.ylabel('True label')\n",
    "    plt.xlabel('Predicted label')\n",
    "\n",
    "# Plot non-normalized confusion matrix\n",
    "plt.figure()\n",
    "plot_confusion_matrix(cnf_matrix, classes=range(6),\n",
    "                      title='Confusion matrix, without normalization')\n",
    "\n",
    "# Plot normalized confusion matrix\n",
    "plt.figure()\n",
    "plot_confusion_matrix(cnf_matrix, classes=range(6), normalize=True,\n",
    "                      title='Normalized confusion matrix')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean accuracy per class\n",
      "0.311391501576\n"
     ]
    }
   ],
   "source": [
    "cm = cnf_matrix.astype('float') / cnf_matrix.sum(axis=1)[:, np.newaxis]\n",
    "print(\"Mean accuracy per class\")\n",
    "print(np.mean(np.diag(cm)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.3049126122499547, 0.31787302098937464)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "proportion_confint(int(0.311391501576 * len(labels_test)), len(labels_test), alpha=0.05, method='wilson')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean accuracy in the correct class or +- 1\n",
      "0.623939074286\n"
     ]
    }
   ],
   "source": [
    "upper = np.diag(cm, 1)\n",
    "upper = np.append(upper, 0)\n",
    "diagonal = np.diag(cm)\n",
    "lower = np.diag(cm, -1)\n",
    "lower = np.append(0, lower)\n",
    "\n",
    "arr = np.vstack((upper, diagonal, lower))\n",
    "print(\"Mean accuracy in the correct class or +- 1\")\n",
    "print(arr.sum(axis=0).mean())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.61713237032946677, 0.63069017823434204)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "proportion_confint(int(0.623939074286 * len(labels_test)), len(labels_test), alpha=0.05, method='wilson')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3D"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.manifold import TSNE\n",
    "x_embedded = TSNE(n_components=3, random_state=1).fit_transform(x_50d)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from mpl_toolkits.mplot3d import Axes3D\n",
    "\n",
    "fig = plt.figure(figsize=(25,25))\n",
    "ax = Axes3D(fig)\n",
    "ax.scatter(x_embedded[:, 0], x_embedded[:, 1], x_embedded[:, 2], c=y, cmap=plt.cm.get_cmap(\"viridis\", 6))\n",
    "plt.savefig('/data/joint_scoring/images/tsne_3d_classification.png', bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already up-to-date: scikit-image in /usr/local/lib/python3.5/dist-packages\n",
      "Requirement already up-to-date: matplotlib>=1.3.1 in /usr/local/lib/python3.5/dist-packages (from scikit-image)\n",
      "Requirement already up-to-date: scipy>=0.17.0 in /usr/local/lib/python3.5/dist-packages (from scikit-image)\n",
      "Requirement already up-to-date: six>=1.7.3 in /usr/local/lib/python3.5/dist-packages (from scikit-image)\n",
      "Requirement already up-to-date: PyWavelets>=0.4.0 in /usr/local/lib/python3.5/dist-packages (from scikit-image)\n",
      "Requirement already up-to-date: pillow>=2.1.0 in /usr/local/lib/python3.5/dist-packages (from scikit-image)\n",
      "Requirement already up-to-date: networkx>=1.8 in /usr/local/lib/python3.5/dist-packages (from scikit-image)\n",
      "Requirement already up-to-date: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.5/dist-packages (from matplotlib>=1.3.1->scikit-image)\n",
      "Requirement already up-to-date: numpy>=1.7.1 in /usr/local/lib/python3.5/dist-packages (from matplotlib>=1.3.1->scikit-image)\n",
      "Requirement already up-to-date: python-dateutil>=2.0 in /usr/local/lib/python3.5/dist-packages (from matplotlib>=1.3.1->scikit-image)\n",
      "Requirement already up-to-date: cycler>=0.10 in /usr/local/lib/python3.5/dist-packages (from matplotlib>=1.3.1->scikit-image)\n",
      "Requirement already up-to-date: pytz in /usr/local/lib/python3.5/dist-packages (from matplotlib>=1.3.1->scikit-image)\n",
      "Requirement already up-to-date: olefile in /usr/local/lib/python3.5/dist-packages (from pillow>=2.1.0->scikit-image)\n",
      "Requirement already up-to-date: decorator>=4.1.0 in /usr/local/lib/python3.5/dist-packages (from networkx>=1.8->scikit-image)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/skimage/transform/_warps.py:84: UserWarning: The default mode, 'constant', will be changed to 'reflect' in skimage 0.15.\n",
      "  warn(\"The default mode, 'constant', will be changed to 'reflect' in \"\n"
     ]
    }
   ],
   "source": [
    "!pip3 install scikit-image --upgrade\n",
    "from skimage.transform import resize\n",
    "imgs_resized = [resize(image=x, output_shape=(50, 50)) for x in imgs]\n",
    "imgs_resized = np.array(imgs_resized)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 68   1  84   1]\n",
      " [ 64  67 135   1]\n",
      " [ 41 120 142   1]\n",
      " [ 34 167 132   1]\n",
      " [121 209  81   1]\n",
      " [253 231  36   1]]\n",
      "<class 'numpy.ndarray'>\n"
     ]
    }
   ],
   "source": [
    "cm = plt.cm.get_cmap(\"viridis\", 6)\n",
    "color = np.ones((6, 4))\n",
    "for i in range(6):\n",
    "    color[i] = cm(1.*i/6)\n",
    "\n",
    "#color = [x[0:-1] for x in color]\n",
    "color = [np.append((x[0:-1] * 255),1) for x in color]\n",
    "color = np.array(color)\n",
    "\n",
    "\n",
    "color = color.astype(\"int16\")\n",
    "\n",
    "print(color)\n",
    "print(type(color[0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(68, 1, 84, 1)"
      ]
     },
     "execution_count": 96,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tuple(color[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "62.845509"
      ]
     },
     "execution_count": 117,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 201,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD8CAYAAABXXhlaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f832ba57208>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(imgs_resized[1].squeeze(), cmap=\"gray\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 221,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Creating images with colored borders, 99.94 % finished\n"
     ]
    }
   ],
   "source": [
    "from PIL import Image\n",
    "from PIL import ImageMath\n",
    "\n",
    "imgs_border = []\n",
    "\n",
    "old_size = (50, 50)\n",
    "new_size = (54, 54)\n",
    "\n",
    "for i, img in enumerate(imgs_resized):\n",
    "    if i % 100 == 0:\n",
    "        clear_output()\n",
    "        print(\"Creating images with colored borders, {0:.2f} % finished\".format(i/19611*100))\n",
    "    old_im = Image.fromarray((img.squeeze()*255).astype(\"int16\"))\n",
    "    old_im = old_im.convert(\"RGB\")\n",
    "    new_im = Image.new(\"RGB\", new_size, tuple(color[int(y[i])]))\n",
    "    new_im.paste(old_im, (2,2))\n",
    "    dat = np.asarray(a=new_im)\n",
    "    imgs_border.append(dat)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 225,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(19611, 54, 54, 3)"
      ]
     },
     "execution_count": 225,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.array(imgs_border).shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 226,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAD8CAYAAABXXhlaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nrkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f83612600b8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(imgs_border[1])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 227,
   "metadata": {},
   "outputs": [],
   "source": [
    "from IPython.display import clear_output\n",
    "res = 8000 # resolution of plot\n",
    "cval = 255 # background colour\n",
    "\n",
    "images = imgs_border\n",
    "max_width = max([image.shape[0] for image in images])\n",
    "max_height = max([image.shape[1] for image in images])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 228,
   "metadata": {},
   "outputs": [],
   "source": [
    "f2d = x_2d\n",
    "\n",
    "xx = f2d[:, 0]\n",
    "yy = f2d[:, 1]\n",
    "x_min, x_max = xx.min(), xx.max()\n",
    "y_min, y_max = yy.min(), yy.max()\n",
    "# Fix the ratios\n",
    "sx = (x_max-x_min)\n",
    "sy = (y_max-y_min)\n",
    "if sx > sy:\n",
    "    res_x = sx/float(sy)*res\n",
    "    res_y = res\n",
    "else:\n",
    "    res_x = res\n",
    "    res_y = sy/float(sx)*res"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 229,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(8000, 8615.1535825786468)"
      ]
     },
     "execution_count": 229,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "res_x, res_y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 230,
   "metadata": {},
   "outputs": [],
   "source": [
    "res_y=res_y.astype(\"int16\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 231,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(54, 54, 3)"
      ]
     },
     "execution_count": 231,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "images[0].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 232,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Creating plot, 99.99 % finished\n"
     ]
    }
   ],
   "source": [
    "canvas = np.ones((res_x+max_width, res_y+max_height, 3))*cval\n",
    "x_coords = np.linspace(x_min, x_max, res_x)\n",
    "y_coords = np.linspace(y_min, y_max, res_y)\n",
    "i = 0\n",
    "for x, y, image in zip(xx, yy, images):\n",
    "    clear_output()\n",
    "    print(\"Creating plot, {0:.2f} % finished\".format(i/19611*100))\n",
    "    i = i + 1\n",
    "    w, h = image.shape[:2]\n",
    "    x_idx = np.argmin((x - x_coords)**2)\n",
    "    y_idx = np.argmin((y - y_coords)**2)\n",
    "    canvas[x_idx:x_idx+w, y_idx:y_idx+h] = image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!rm /data/joint_scoring/images/embeddings_classification.png"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 234,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.figure(figsize=(400,400))\n",
    "plt.imshow(canvas)\n",
    "plt.savefig('/data/joint_scoring/images/embeddings_classification.png', bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}