2-3-HPF parameter tuning on validation set.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import datetime\n",
    "import os\n",
    "\n",
    "import pandas as pd\n",
    "\n",
    "from src.config import DATA_DIR, MODEL_DIR, PARAM_DIR\n",
    "from src.recommendation.HPF import train_hpf_model, model_type\n",
    "from src.recommendation.all_evaluation import _results\n",
    "from src.recommendation.evaluation import evaluate\n",
    "from src.recommendation.mixture_functions import get_train\n",
    "from src.util.data_io import get_hpf_dataset\n",
    "from src.util.eval_measures import eval_lst\n",
    "from src.util.io import save_pickle\n",
    "\n",
    "train_model, model_type = train_hpf_model, model_type\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'hpf_model'"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_type\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_dir = os.path.join(DATA_DIR, 'day')\n",
    "results_dir = os.path.join(MODEL_DIR, 'tuning')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[['recall', [5]], ['precision', [5]], ['nDCG', [5]]]\n"
     ]
    }
   ],
   "source": [
    "print (eval_lst)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# subset Jan 20 - 22, 2015\n",
    "datasets = ['120', '121', '122']\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "k_vals = [5, 10, 50, 100, 500]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "overwrite = False\n",
    "save_multinomials = True\n",
    "\n",
    "\n",
    "def hours_minutes_seconds(td):\n",
    "    return '{:02}:{:02}:{:02}'.format(td.seconds // 3600, (td.seconds // 60) % 60, td.seconds % 60)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def eval_dataset(dataset, k_vals):\n",
    "    result_dic = {}\n",
    "\n",
    "    for n_components in k_vals:\n",
    "        print(n_components)\n",
    "        train, val, test = get_hpf_dataset(dataset, data_dir)\n",
    "        start_time = datetime.datetime.now()\n",
    "        all_multinomials = train_model(train, val, test, results_dir,\n",
    "                                       dataset, overwrite=False, save_multinomials=True,\n",
    "                                       n_components=n_components, top_n=10)\n",
    "        end_time = datetime.datetime.now()\n",
    "        time_used = hours_minutes_seconds(end_time - start_time)\n",
    "        print('time_used', time_used)\n",
    "        dataset, eval_result = _results(dataset, eval_lst, val.values, all_multinomials)\n",
    "        eval_result.append(['time_used', time_used])\n",
    "        result_dic.update({n_components: eval_result})\n",
    "\n",
    "    df = pd.DataFrame(result_dic).T\n",
    "    df.columns = [item for sublist in list(result_dic.values())[0] for item in sublist[:1]]\n",
    "    for col in df.columns:\n",
    "        df[col] = df[col].apply(lambda s: s[1])\n",
    "\n",
    "    return df\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5\n",
      "2019-10-22 17:38:17,877 INFO --> Loading data for 120\n",
      "time_used 00:00:03\n",
      "recall@5: 0.04903\n",
      "precision@5: 0.05507\n",
      "nDCG@5: 0.07063\n",
      "10\n",
      "2019-10-22 17:38:59,098 INFO --> Loading data for 120\n",
      "**********************************\n",
      "Hierarchical Poisson Factorization\n",
      "**********************************\n",
      "\n",
      "Number of users: 3432\n",
      "Number of items: 29474\n",
      "Latent factors to use: 10\n",
      "\n",
      "Initializing parameters...\n",
      "Allocating Phi matrix...\n",
      "Initializing optimization procedure...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/yueliu/anaconda3/lib/python3.6/site-packages/hpfrec/__init__.py:539: UserWarning: 'val_set' contains observations with a count value less than 1, these will be ignored.\n",
      "  warnings.warn(msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration 10 | val llk: -102856 | val rmse: 0.9649\n",
      "Iteration 20 | val llk: -101928 | val rmse: 0.9644\n",
      "Iteration 30 | val llk: -101602 | val rmse: 0.9642\n",
      "Iteration 40 | val llk: -101500 | val rmse: 0.9641\n",
      "Iteration 50 | val llk: -101438 | val rmse: 0.9641\n",
      "\n",
      "\n",
      "Optimization finished\n",
      "Final log-likelihood: -101438\n",
      "Final RMSE: 0.9641\n",
      "Minutes taken (optimization part): 0.1\n",
      "\n",
      "time_used 00:06:09\n",
      "recall@5: 0.05086\n",
      "precision@5: 0.05647\n",
      "nDCG@5: 0.07311\n",
      "50\n",
      "2019-10-22 17:45:45,184 INFO --> Loading data for 120\n",
      "**********************************\n",
      "Hierarchical Poisson Factorization\n",
      "**********************************\n",
      "\n",
      "Number of users: 3432\n",
      "Number of items: 29474\n",
      "Latent factors to use: 50\n",
      "\n",
      "Initializing parameters...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/yueliu/anaconda3/lib/python3.6/site-packages/hpfrec/__init__.py:539: UserWarning: 'val_set' contains observations with a count value less than 1, these will be ignored.\n",
      "  warnings.warn(msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Allocating Phi matrix...\n",
      "Initializing optimization procedure...\n",
      "Iteration 10 | val llk: -93986 | val rmse: 1.0031\n",
      "Iteration 20 | val llk: -92832 | val rmse: 1.0581\n",
      "Iteration 30 | val llk: -92714 | val rmse: 1.0996\n",
      "Iteration 40 | val llk: -92665 | val rmse: 1.1131\n",
      "\n",
      "\n",
      "Optimization finished\n",
      "Final log-likelihood: -92665\n",
      "Final RMSE: 1.1131\n",
      "Minutes taken (optimization part): 0.4\n",
      "\n",
      "time_used 00:05:47\n",
      "recall@5: 0.06694\n",
      "precision@5: 0.07512\n",
      "nDCG@5: 0.09899\n",
      "100\n",
      "2019-10-22 17:52:09,788 INFO --> Loading data for 120\n",
      "**********************************\n",
      "Hierarchical Poisson Factorization\n",
      "**********************************\n",
      "\n",
      "Number of users: 3432\n",
      "Number of items: 29474\n",
      "Latent factors to use: 100\n",
      "\n",
      "Initializing parameters...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/yueliu/anaconda3/lib/python3.6/site-packages/hpfrec/__init__.py:539: UserWarning: 'val_set' contains observations with a count value less than 1, these will be ignored.\n",
      "  warnings.warn(msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Allocating Phi matrix...\n",
      "Initializing optimization procedure...\n",
      "Iteration 10 | val llk: -89976 | val rmse: 1.0294\n",
      "Iteration 20 | val llk: -89373 | val rmse: 1.1345\n",
      "Iteration 30 | val llk: -89504 | val rmse: 1.1665\n",
      "\n",
      "\n",
      "Optimization finished\n",
      "Final log-likelihood: -89504\n",
      "Final RMSE: 1.1665\n",
      "Minutes taken (optimization part): 0.4\n",
      "\n",
      "time_used 00:05:46\n",
      "recall@5: 0.08155\n",
      "precision@5: 0.09062\n",
      "nDCG@5: 0.11894\n",
      "500\n",
      "2019-10-22 17:58:32,984 INFO --> Loading data for 120\n",
      "**********************************\n",
      "Hierarchical Poisson Factorization\n",
      "**********************************\n",
      "\n",
      "Number of users: 3432\n",
      "Number of items: 29474\n",
      "Latent factors to use: 500\n",
      "\n",
      "Initializing parameters...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/yueliu/anaconda3/lib/python3.6/site-packages/hpfrec/__init__.py:539: UserWarning: 'val_set' contains observations with a count value less than 1, these will be ignored.\n",
      "  warnings.warn(msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Allocating Phi matrix...\n",
      "Initializing optimization procedure...\n",
      "Iteration 10 | val llk: -81360 | val rmse: 1.2401\n",
      "Iteration 20 | val llk: -81868 | val rmse: 1.2640\n",
      "\n",
      "\n",
      "Optimization finished\n",
      "Final log-likelihood: -81868\n",
      "Final RMSE: 1.2640\n",
      "Minutes taken (optimization part): 0.8\n",
      "\n",
      "time_used 00:05:11\n",
      "recall@5: 0.09969\n",
      "precision@5: 0.11259\n",
      "nDCG@5: 0.15027\n",
      "5\n",
      "2019-10-22 18:04:21,898 INFO --> Loading data for 121\n",
      "**********************************\n",
      "Hierarchical Poisson Factorization\n",
      "**********************************\n",
      "\n",
      "Number of users: 3641\n",
      "Number of items: 30195\n",
      "Latent factors to use: 5\n",
      "\n",
      "Initializing parameters...\n",
      "Allocating Phi matrix...\n",
      "Initializing optimization procedure...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/yueliu/anaconda3/lib/python3.6/site-packages/hpfrec/__init__.py:539: UserWarning: 'val_set' contains observations with a count value less than 1, these will be ignored.\n",
      "  warnings.warn(msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration 10 | val llk: -118464 | val rmse: 0.9676\n",
      "Iteration 20 | val llk: -117897 | val rmse: 0.9672\n",
      "Iteration 30 | val llk: -117675 | val rmse: 0.9672\n",
      "Iteration 40 | val llk: -117627 | val rmse: 0.9673\n",
      "\n",
      "\n",
      "Optimization finished\n",
      "Final log-likelihood: -117627\n",
      "Final RMSE: 0.9673\n",
      "Minutes taken (optimization part): 0.1\n",
      "\n",
      "time_used 00:04:04\n",
      "recall@5: 0.04959\n",
      "precision@5: 0.05581\n",
      "nDCG@5: 0.07323\n",
      "10\n",
      "2019-10-22 18:09:07,519 INFO --> Loading data for 121\n",
      "**********************************\n",
      "Hierarchical Poisson Factorization\n",
      "**********************************\n",
      "\n",
      "Number of users: 3641\n",
      "Number of items: 30195\n",
      "Latent factors to use: 10\n",
      "\n",
      "Initializing parameters...\n",
      "Allocating Phi matrix...\n",
      "Initializing optimization procedure...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/yueliu/anaconda3/lib/python3.6/site-packages/hpfrec/__init__.py:539: UserWarning: 'val_set' contains observations with a count value less than 1, these will be ignored.\n",
      "  warnings.warn(msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration 10 | val llk: -114798 | val rmse: 0.9709\n",
      "Iteration 20 | val llk: -113752 | val rmse: 0.9753\n",
      "Iteration 30 | val llk: -113430 | val rmse: 0.9776\n",
      "Iteration 40 | val llk: -113329 | val rmse: 0.9791\n",
      "\n",
      "\n",
      "Optimization finished\n",
      "Final log-likelihood: -113329\n",
      "Final RMSE: 0.9791\n",
      "Minutes taken (optimization part): 0.1\n",
      "\n",
      "time_used 00:04:16\n",
      "recall@5: 0.05243\n",
      "precision@5: 0.05932\n",
      "nDCG@5: 0.07750\n",
      "50\n",
      "2019-10-22 18:14:03,222 INFO --> Loading data for 121\n",
      "**********************************\n",
      "Hierarchical Poisson Factorization\n",
      "**********************************\n",
      "\n",
      "Number of users: 3641\n",
      "Number of items: 30195\n",
      "Latent factors to use: 50\n",
      "\n",
      "Initializing parameters...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/yueliu/anaconda3/lib/python3.6/site-packages/hpfrec/__init__.py:539: UserWarning: 'val_set' contains observations with a count value less than 1, these will be ignored.\n",
      "  warnings.warn(msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Allocating Phi matrix...\n",
      "Initializing optimization procedure...\n",
      "Iteration 10 | val llk: -104641 | val rmse: 0.9830\n",
      "Iteration 20 | val llk: -103173 | val rmse: 1.0116\n",
      "Iteration 30 | val llk: -102910 | val rmse: 1.0332\n",
      "Iteration 40 | val llk: -102878 | val rmse: 1.0576\n",
      "\n",
      "\n",
      "Optimization finished\n",
      "Final log-likelihood: -102878\n",
      "Final RMSE: 1.0576\n",
      "Minutes taken (optimization part): 0.3\n",
      "\n",
      "time_used 00:05:09\n",
      "recall@5: 0.06778\n",
      "precision@5: 0.07696\n",
      "nDCG@5: 0.10014\n",
      "100\n",
      "2019-10-22 18:19:52,553 INFO --> Loading data for 121\n",
      "**********************************\n",
      "Hierarchical Poisson Factorization\n",
      "**********************************\n",
      "\n",
      "Number of users: 3641\n",
      "Number of items: 30195\n",
      "Latent factors to use: 100\n",
      "\n",
      "Initializing parameters...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/yueliu/anaconda3/lib/python3.6/site-packages/hpfrec/__init__.py:539: UserWarning: 'val_set' contains observations with a count value less than 1, these will be ignored.\n",
      "  warnings.warn(msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Allocating Phi matrix...\n",
      "Initializing optimization procedure...\n",
      "Iteration 10 | val llk: -100273 | val rmse: 1.0116\n",
      "Iteration 20 | val llk: -99399 | val rmse: 1.1155\n",
      "Iteration 30 | val llk: -99533 | val rmse: 1.1614\n",
      "\n",
      "\n",
      "Optimization finished\n",
      "Final log-likelihood: -99533\n",
      "Final RMSE: 1.1614\n",
      "Minutes taken (optimization part): 0.4\n",
      "\n",
      "time_used 00:04:45\n",
      "recall@5: 0.08937\n",
      "precision@5: 0.10426\n",
      "nDCG@5: 0.13334\n",
      "500\n",
      "2019-10-22 18:25:19,766 INFO --> Loading data for 121\n",
      "**********************************\n",
      "Hierarchical Poisson Factorization\n",
      "**********************************\n",
      "\n",
      "Number of users: 3641\n",
      "Number of items: 30195\n",
      "Latent factors to use: 500\n",
      "\n",
      "Initializing parameters...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/yueliu/anaconda3/lib/python3.6/site-packages/hpfrec/__init__.py:539: UserWarning: 'val_set' contains observations with a count value less than 1, these will be ignored.\n",
      "  warnings.warn(msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Allocating Phi matrix...\n",
      "Initializing optimization procedure...\n",
      "Iteration 10 | val llk: -90386 | val rmse: 1.0888\n",
      "Iteration 20 | val llk: -91086 | val rmse: 1.1760\n",
      "\n",
      "\n",
      "Optimization finished\n",
      "Final log-likelihood: -91086\n",
      "Final RMSE: 1.1760\n",
      "Minutes taken (optimization part): 0.7\n",
      "\n",
      "time_used 00:04:51\n",
      "recall@5: 0.10102\n",
      "precision@5: 0.11903\n",
      "nDCG@5: 0.15303\n",
      "5\n",
      "2019-10-22 18:30:51,317 INFO --> Loading data for 122\n",
      "**********************************\n",
      "Hierarchical Poisson Factorization\n",
      "**********************************\n",
      "\n",
      "Number of users: 3591\n",
      "Number of items: 30106\n",
      "Latent factors to use: 5\n",
      "\n",
      "Initializing parameters...\n",
      "Allocating Phi matrix...\n",
      "Initializing optimization procedure...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/yueliu/anaconda3/lib/python3.6/site-packages/hpfrec/__init__.py:539: UserWarning: 'val_set' contains observations with a count value less than 1, these will be ignored.\n",
      "  warnings.warn(msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration 10 | val llk: -116322 | val rmse: 0.9651\n",
      "Iteration 20 | val llk: -115604 | val rmse: 0.9647\n",
      "Iteration 30 | val llk: -115428 | val rmse: 0.9646\n",
      "Iteration 40 | val llk: -115375 | val rmse: 0.9645\n",
      "\n",
      "\n",
      "Optimization finished\n",
      "Final log-likelihood: -115375\n",
      "Final RMSE: 0.9645\n",
      "Minutes taken (optimization part): 0.1\n",
      "\n",
      "time_used 00:03:52\n",
      "recall@5: 0.05094\n",
      "precision@5: 0.05803\n",
      "nDCG@5: 0.07483\n",
      "10\n",
      "2019-10-22 18:35:22,336 INFO --> Loading data for 122\n",
      "**********************************\n",
      "Hierarchical Poisson Factorization\n",
      "**********************************\n",
      "\n",
      "Number of users: 3591\n",
      "Number of items: 30106\n",
      "Latent factors to use: 10\n",
      "\n",
      "Initializing parameters...\n",
      "Allocating Phi matrix...\n",
      "Initializing optimization procedure...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/yueliu/anaconda3/lib/python3.6/site-packages/hpfrec/__init__.py:539: UserWarning: 'val_set' contains observations with a count value less than 1, these will be ignored.\n",
      "  warnings.warn(msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration 10 | val llk: -112782 | val rmse: 0.9739\n",
      "Iteration 20 | val llk: -111599 | val rmse: 0.9871\n",
      "Iteration 30 | val llk: -111231 | val rmse: 0.9979\n",
      "Iteration 40 | val llk: -111130 | val rmse: 1.0061\n",
      "\n",
      "\n",
      "Optimization finished\n",
      "Final log-likelihood: -111130\n",
      "Final RMSE: 1.0061\n",
      "Minutes taken (optimization part): 0.1\n",
      "\n",
      "time_used 00:03:53\n",
      "recall@5: 0.05508\n",
      "precision@5: 0.06321\n",
      "nDCG@5: 0.08271\n",
      "50\n",
      "2019-10-22 18:39:56,541 INFO --> Loading data for 122\n",
      "**********************************\n",
      "Hierarchical Poisson Factorization\n",
      "**********************************\n",
      "\n",
      "Number of users: 3591\n",
      "Number of items: 30106\n",
      "Latent factors to use: 50\n",
      "\n",
      "Initializing parameters...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/yueliu/anaconda3/lib/python3.6/site-packages/hpfrec/__init__.py:539: UserWarning: 'val_set' contains observations with a count value less than 1, these will be ignored.\n",
      "  warnings.warn(msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Allocating Phi matrix...\n",
      "Initializing optimization procedure...\n",
      "Iteration 10 | val llk: -102459 | val rmse: 0.9774\n",
      "Iteration 20 | val llk: -100989 | val rmse: 1.0128\n",
      "Iteration 30 | val llk: -100785 | val rmse: 1.0515\n",
      "Iteration 40 | val llk: -100748 | val rmse: 1.0814\n",
      "\n",
      "\n",
      "Optimization finished\n",
      "Final log-likelihood: -100748\n",
      "Final RMSE: 1.0814\n",
      "Minutes taken (optimization part): 0.3\n",
      "\n",
      "time_used 00:04:18\n",
      "recall@5: 0.07307\n",
      "precision@5: 0.08248\n",
      "nDCG@5: 0.10925\n",
      "100\n",
      "2019-10-22 18:44:55,598 INFO --> Loading data for 122\n",
      "**********************************\n",
      "Hierarchical Poisson Factorization\n",
      "**********************************\n",
      "\n",
      "Number of users: 3591\n",
      "Number of items: 30106\n",
      "Latent factors to use: 100\n",
      "\n",
      "Initializing parameters...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/yueliu/anaconda3/lib/python3.6/site-packages/hpfrec/__init__.py:539: UserWarning: 'val_set' contains observations with a count value less than 1, these will be ignored.\n",
      "  warnings.warn(msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Allocating Phi matrix...\n",
      "Initializing optimization procedure...\n",
      "Iteration 10 | val llk: -98448 | val rmse: 1.0398\n",
      "Iteration 20 | val llk: -97545 | val rmse: 1.1141\n",
      "Iteration 30 | val llk: -97615 | val rmse: 1.1400\n",
      "\n",
      "\n",
      "Optimization finished\n",
      "Final log-likelihood: -97615\n",
      "Final RMSE: 1.1400\n",
      "Minutes taken (optimization part): 0.3\n",
      "\n",
      "time_used 00:04:16\n",
      "recall@5: 0.08746\n",
      "precision@5: 0.10114\n",
      "nDCG@5: 0.12982\n",
      "500\n",
      "2019-10-22 18:49:52,681 INFO --> Loading data for 122\n",
      "**********************************\n",
      "Hierarchical Poisson Factorization\n",
      "**********************************\n",
      "\n",
      "Number of users: 3591\n",
      "Number of items: 30106\n",
      "Latent factors to use: 500\n",
      "\n",
      "Initializing parameters...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/yueliu/anaconda3/lib/python3.6/site-packages/hpfrec/__init__.py:539: UserWarning: 'val_set' contains observations with a count value less than 1, these will be ignored.\n",
      "  warnings.warn(msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Allocating Phi matrix...\n",
      "Initializing optimization procedure...\n",
      "Iteration 10 | val llk: -89140 | val rmse: 1.2012\n",
      "Iteration 20 | val llk: -89676 | val rmse: 1.2427\n",
      "\n",
      "\n",
      "Optimization finished\n",
      "Final log-likelihood: -89676\n",
      "Final RMSE: 1.2427\n",
      "Minutes taken (optimization part): 0.7\n",
      "\n",
      "time_used 00:05:03\n",
      "recall@5: 0.10674\n",
      "precision@5: 0.12671\n",
      "nDCG@5: 0.16704\n"
     ]
    }
   ],
   "source": [
    "dfs = dict()\n",
    "for dataset in datasets:\n",
    "    dataset = str(dataset)\n",
    "    df = eval_dataset(dataset, k_vals)\n",
    "    dfs.update({dataset: df})\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['recall@5', 'precision@5', 'nDCG@5', 'time_used']"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "eval_measure = df.columns.tolist()\n",
    "eval_measure\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_metric(col):\n",
    "    try:\n",
    "        df_plot = pd.DataFrame({c: df[col] for c, df in dfs.items()})\n",
    "        df_plot.plot(logx=True, xticks=k_vals, title=col.upper() + ' for different k')\n",
    "    except:\n",
    "        # time is not numeric value\n",
    "        return df_plot\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "120\n",
      "500\n",
      "121\n",
      "500\n",
      "122\n",
      "500\n"
     ]
    }
   ],
   "source": [
    "for k, df in dfs.items():\n",
    "    print(k)\n",
    "    print(df['nDCG@5'].idxmax())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "          120       121       122\n",
      "5    00:00:03  00:04:04  00:03:52\n",
      "10   00:06:09  00:04:16  00:03:53\n",
      "50   00:05:47  00:05:09  00:04:18\n",
      "100  00:05:46  00:04:45  00:04:16\n",
      "500  00:05:11  00:04:51  00:05:03\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAEMCAYAAAAoB2Y1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAIABJREFUeJzs3Xdc1WX/x/HXxRZxITiR4URxi3uv1KZmw2yZlnX30/a8687SSi3vxp3dd2maq+3KhhNnWiluhgNBposhIAgczrl+f3yPhoZwUOAwPs/Hg0dwzndcx+B9rnNd3+/nUlprhBBCVA8O9m6AEEKI8iOhL4QQ1YiEvhBCVCMS+kIIUY1I6AshRDUioS+EENWIhL4QQlQjEvqi2lNKNVRKbVdKZSql/l0Gx/dXSmmllJP157VKqYcLPP+2UipZKXXa+vMYpVS8UuqCUqpLabfnRimlTiqlhtm7HeL6SOiLQln/sM8opWoWeOxRpdTWAj9rpVSWNZxSlFIhSql7CznWiAKhek4ptU0pdXuB5xsrpeYrpZKsx4pWSi1SSgVedRwnpdTjSqmd1pBMVEqtVkoNLuScBdt2QSn1RREvdzKQDNTWWj9fsn+pktNaj9JaL7a2sxnwPNBOa93IuskcYIrW2kNrvb+s21OQUupNpdSy8jynKF8S+qIoTsDTxWzTSWvtAbQBFgFzlVLTLj2plLoL+AFYAvgADYE3gNusz9cHdgHuQH+gFtAV2AYML3CcmsBGYBhGSDYDWgD/BWYppd64VtusX48W8Rr8gAh9HbenX+q93wA/IEVrffaqx8Kv52Cl0B5R1Wmt5Uu+/vYFnAReAVKButbHHgW2FthGAy2v2u8uIAeoDyggDnixiPO8DRwEHIppz0Jg+jWecwV2AAOKats19l0EmIA84ALGm4or8BGQZP36CHC1bj8ISABeBk4DSws5piNGbz0ZiAb+z9oeJ+vzW63/lsOAi4DFeu5vrP/VQBZwwrp9E2AFcA6IAZ4qcK43geXAMiDDelwH6/+7E0AK8D3gad3e33r8h63/b5KB16zPjbT+O5is7ThYxO/GMOv3gdY2jbP376x82fZl9wbIV8X8uvSHDawE3rY+ZkvoOwP5wChrIGggoIjz/AG8WUxbAoAwa5jVBVZZA/db4CeMnnFfYPlVbUuybrcS8C/i+IsuvUbrz9Ot7WoAeGN8EplhfW6Q9fXNxnhzqFHI8Z4AjmB8GvEEthQW+gWOl3DV/pf/Xa2veS/GpyMXoDnGG8kI6/NvWkN6tHXbGsAz1vb7WNv4OfCNdftLoT/fum0nIBdoW+B4y2z83eiK8cZxq71/X+XL9i8Z3hHFeQOYqpTytmVjrbUJo/foidHbBzhVxC5eGMEMgFLqdqXUeev4/wbrw0OBFVprC/AqRo/WD3gfGILxieIAxpvMJQMxAi4QI/x/LsHQx/0YnyrOaq3PAW8BDxZ43gJM01rnaq0vFrL/PcBHWut4rXUqMNPG8xamO+CttZ6utc7TWkdjBPa4Atv8rrVerbW2WNvzOEbvPUFrnYsR5Hdd9frf0lpf1FofxPik1amE7eoPrAEe1lr/fJ2vTdiBjP+JImmtw5RSP2MMF0QWt71Syhmjd5yKMbQA0BhjCKAwKdbnL51vDVBXKfUo8ID14QZAovX7DsBca5jtVUpdGvtuVmAbtNbbrd/mKaWexnijaAscLu41YAynxBb4Odb62CXntNY5xewff9X+18sPaKKUOl/gMUeM4axL4q/cBT9glVLKUuAxM8Z8yiWnC3yfDXiUsF1PANu01ltKuJ+wM+npC1tMAx4Dmtqw7R0Ywx+7gaMYgTS2iO1DgNFKqaJ+F5P5643hMHCPUspVKdUVCLI+9wnGMMa1aIxPBLZIwgjOS3ytjxU8VlFOYbwJFdz/esUDMVrrugW+ammtby6iPfHAqKv2cdNaJ1I8WyeznwB8lVIf2ri9qCAk9EWxtNZRwHfAU9faRinlqZS6H/gUmK21TtFaa+A54F9KqUeUUrWVUg5KqX5KqXnWXT8A6gFLlVItlKEW0LnA4bcAY5RSCmOopA7GWPIbwHqMidb5WuuV1rYEKaU6K6UclVIewL8xPgUU+0nF6hvgdaWUt1LKy3qeklzG+D3wlFLKRylVD+NT0vXaDWQopV5WStWwvqb2SqnuRezzGfCOUsoPwPo67rDxfGcA/2LehAEyMSZ+ByilZtl4bFEBSOgLW00Hahby+EGl1AUgCmOi91mt9eXLJ7XWy4F7gYkYveUzGFfs/Gh9PhnohXHFz28YYXIA49LNf1i3OQ5EAK9qrc9rrcdorRtqrUdrre8E+mqtvy/QpoYYb1IZGJOe/hiTjSYbX+vbQChwCOOTxT7rY7aaj/FmdNC678oS7HsFrbUZ4/LWzhhDZMnAFxhvfNfyMcZ4+walVCbGpG5PG0/5g/W/KUqpfcW07TzGZbWjlFIzbDy+sDNldMaEqNiUUrWBdRi99Q+t/60HPALcBwRbJ3qFEEWQnr6oFLTWGcBgjJuWFmP0ePcBLYExEvhC2EZ6+kIIUY1IT18IIaoRCX0hhKhGKtzNWV5eXtrf39/ezRBCiEpl7969yVrrYu+cr3Ch7+/vT2hoqL2bIYQQlYpSyqY7v2V4RwghqhEJfSGEqEYk9IUQohqpcGP6hTGZTCQkJJCTU1Rhw8rJzc0NHx8fnJ2d7d0UIUQ1UClCPyEhgVq1auHv749Rc6tq0FqTkpJCQkICAQEB9m6OEKIaqBTDOzk5OdSvX79KBT6AUor69etXyU8wQoiKqVKEPlDlAv+Sqvq6hBDlJz033eZtK03oVwQTJ06kQYMGtG/f/vJjL774IoGBgXTs2JExY8Zw/vxfCxzNnDmTli1b0qZNG9avX2+PJgshqriQ2BBG/zja5u0l9EtgwoQJrFu37orHhg8fTlhYGIcOHaJ169bMnGkshxoREcG3335LeHg469at48knn8RsNtuj2UKIKij5YjLPb32eZ7Y+Q323+sXvYCWhXwIDBgzA09PzisduuukmnJyM+fBevXqRkJAAwI8//si4ceNwdXUlICCAli1bsnv37nJvsxCiatFa89OJnxj942i2xG9hapepfHPrNzbvXymu3inorZ/CiUjKKNVjtmtSm2m3Bd3wcRYuXMi9994LQGJiIr169br8nI+PD4mJtixRKoQQhTuddZoZf8xge8J2Onp3ZHqf6bSo26JEx6h0oV9RvfPOOzg5OXH//fcDxrvx1WTSVghxPSzawvJjy/lg7wdYtIWXur/E+MDxODo4lvhYlS70S6NHXtoWL17Mzz//TEhIyOVg9/HxIT4+/vI2CQkJNGnSxF5NFEJUUvEZ8Uz7fRp7Tu+hZ6OeTOszjWa1ml338WRM/watW7eO2bNns2bNGtzd3S8/fvvtt/Ptt9+Sm5tLTEwMx48fp0ePHnZsqRCiMjFbzCwOX8yda+4kMiWSab2nMf+m+TcU+FAJe/r2dN9997F161aSk5Px8fHhrbfeYubMmeTm5jJ8+HDAmMz97LPPCAoK4p577qFdu3Y4OTnx6aef4uhY8o9iQojqJyotimm7pnEo+RADfQbyeq/XaVSzUakcu8KtkRscHKyvrqcfGRlJ27Zt7dSislfVX58QwjYms4kFYQv4/NDneDh78EqPV7g54Gab5gOVUnu11sHFbSc9fSGEqADCU8J5Y+cbHEs7xkj/kbzS4xXq17D9+ntbSegLIYQd5eTn8L+D/2Nx+GI83Tz5ePDHDPEdUmbnk9AXQgg72XdmH9N2TeNkxknGtBzDC91foLZL7TI9p01X7yilRiqljiqlopRSrxTy/ACl1D6lVL5S6q6rnvNVSm1QSkUqpSKUUv6l03QhhKicskxZvPvnu0xYNwGTxcTnwz9net/pZR74YENPXynlCHwKDAcSgD1KqTVa64gCm8UBE4AXCjnEEuAdrfVGpZQHYLnhVgshRCW1K3EXb/7+JqezTjO+7Xie6vIU7s7uxe9YSmwZ3ukBRGmtowGUUt8CdwCXQ19rfdL63BWBrpRqBzhprTdat7tQOs0WQojKJT03nff3vM+PJ37Ev7Y/i0ctpkuDLuXeDluGd5oC8QV+TrA+ZovWwHml1Eql1H6l1PvWTw5XUEpNVkqFKqVCz507Z+Ohy19JSiunpKQwePBgPDw8mDJlir2aLISoAC6VP/45+mce7fAoy29fbpfAB9tCv7ALRG29uN8J6I8x7NMdaI4xDHTlwbSep7UO1loHe3t723jo8leS0spubm7MmDGDOXPm2KOpQogK4Oryx1/f8jVPd30aV0dXu7XJltBPAAre9+sDJNl4/ARgv9Y6WmudD6wGupasiRVHSUor16xZk379+uHm5lbu7RRC2Ne1yh+3q9/O3k2zaUx/D9BKKRUAJALjgPE2Hn8PUE8p5a21PgcMAUKL2adoa1+B04dv6BB/06gDjJp1w4cpWFpZCFE9lUb547JUbOhrrfOVUlOA9YAjsFBrHa6Umg6Eaq3XKKW6A6uAesBtSqm3tNZBWmuzUuoFIEQZ9xHvBeaX3cuxn6tLKwshqpfSLH9clmy6OUtr/Svw61WPvVHg+z0Ywz6F7bsR6HgDbbxSKfTIS1thpZWFENVHaZc/LktyR+4NulRaedu2bVeUVhZCVH1mi5llkcuYu38uTg5OTOs9jbGtxlbozp+EfgmUpLQygL+/PxkZGeTl5bF69Wo2bNhAu3b2n8gRQty4six/XJYk9Evgm2/+vvjwpEmTrrn9yZMny7A1Qgh7MFlMLDj8V/njWf1n2Vz+uCKQ0BdCCBuVV/njsiShL4QQxSjv8sdlSUJfCCGKYI/yx2VJQl8IIQqRbcrmo30f8e2Rb2ni0YTPh39OnyZ97N2sGyahL4QQV9mVuIu3fn+LU1mnuC/wPp7u+nS5lj8ukdxMOLrW5s0l9IUQwio9N505oXNYHbXaruWPi2W6CMc3QNgKOLYe8nNs3tWmlbOEoSSllTdu3Ei3bt3o0KED3bp1Y/PmzfZqthDCBiFxRvnjn078ZPfyx4Uym+DYBlj5OLzfCr5/CGJ3QdeHYOJ6mw8jPf0SmDBhAlOmTOGhhx66/Njw4cOZOXMmTk5OvPzyy8ycOZPZs2fj5eXFTz/9RJMmTQgLC2PEiBEkJibasfVCiMKkXExh5u6ZrD+5njb12vDp0E8rRDVMACxmiN1p9OgjfoSLaeBWB4JGQ/ux4N8fHEsW4xL6JTBgwIC/3XB10003Xf6+V69eLF++HIAuXf7qIQQFBZGTk0Nubi6urvaroy2E+IvWml9ifmH27tlkmbKY2mUqj7R/BGcHZ3s3DBJCjaAPXwUXToNzTQi82Qj6FkPByeW6D1/pQn/27tkcST1SqscM9Azk5R4v3/BxrlVaecWKFXTp0kUCX4gKosKVP9YazoQZQR+2As7HgaMrtBpuBH3rkeBSOhPJlS70K6prlVYODw/n5ZdfZsOGDXZqmRDiEou2sOL4Cj4I/YB8S779yx8nR1mDfjkkHwPlCC0Gw6BXIfAWYyinGKfTc/hw4zGbT1npQr80euSl7VqllRMSEhgzZgxLliyhRYuKs4iCENVRhSl/fD4ewlfC4eVw+hCgwK8v9HwC2o2GmraVdcjMMfH5tmi++C0as8XWFWwrYehXNNcqrXz+/HluueUWZs6cSd++fe3YQiGqtwpR/vjCWQhfbfTo4/80HmvaDUbMNCZlazex+VAms4Vvdsfx8abjpGTlcXunJrw4og2+79q2v4R+CZSktPLcuXOJiopixowZzJgxA4ANGzbQoEEDe74EIaoVu5Y/vpgGkT8ZPfqTO0BboEEQDH0Dgu4Ez4ASHU5rzfrw08xed5SY5Cx6Bniy8Oa2dGpWt0THUVrb/rGgPAQHB+vQ0CuX0Y2MjKRt27Z2alHZq+qvT4jydqn88bxD86jpXJNXerxSPuWPcy8Yd8eGLYeoELCYwLM5tL8L2t8JDa7v73xvbCrv/nqEvbFptGzgwaujAhkS2OCK16OU2qu1Di7uWNLTF0JUKeVe/tiUA1EbjR79sfWQfxFqN4VeTxhX3jTuDNf5ZhOTnMV7646wNuw03rVcmXlnB+7u5oOT4/XfVyuhL4SoEsq1/LHZBNHbjB595M+QlwnuXtDlASPom/UEh+sP5pQLufwn5Dhf/RmHi5MDzw5rzWMDAnB3ufHIltAXQlR65VL+2GKGuN+NHn3Ej3AxFVzrQNAd1rtjB5T47tirXcwzs3BnDP/beoKLJjPjujfjmWGt8a5Vevf4VJrQ11pXmuXISqKizakIUZmUefljrSFxn9GjD18FmafA2R3ajDLG6VsOBacbD2SzRbNiXwIfbDjG6YwchrdryMsjA2nZwKMUXsSVKkXou7m5kZKSQv369atU8GutSUlJwc3Nzd5NEaLSKbPyx1rDmfACd8fGgqMLtBwOHS7dHVvzxs+DkQHbjp1j1tojHDmdSadmdfl4XGd6Ni+7OYhKEfo+Pj4kJCRw7tw5ezel1Lm5ueHj42PvZghRaZRZ+eOUE38F/bkjxt2xzQfCwJeNu2NrlOzSyOKEJaYzc20kO6NS8PV0Z+74LtzSoXGZd2wrReg7OzsTEFCya1qFEFVPSFwIb//xNmk5aTza4VGe6PQEro43MLySngBhK42gP3XAeMy3D9zyb2h7B3h4l07DC0hIy+aDDcdYdSCRujWcmXZbO+7v6YeLU/lUuq8UoS+EqN5KtfzxhbPGRGzYCmNiFqBJF7jpHQgaA3Wall7DC0i/aOK/W6L4ctdJAB4f0IJ/DGpBnRrlW9VTQl8IUWGVWvnji2nGpZVhKyBmm3F3rHdbGPK6cXds/bKrjZWbb2bZH3F8svk46RdNjOnSlOdvakPTujXK7JxFkdAXQlRIN1z+OPcCHFtnBP3xjcbdsfX8od9zxiWWDct2oRStNT8dOsX7648Qn3qR/q28eGVUIEFNiq+cWZYk9IUQFcoNlT825UDUJuvasevAlA21mkDPx40yCE26XvfdsSXxR3QKM3+N5GBCOoGNarFkYg8GtC79+YHrIaEvhKgwrqv8sdlkDNmErTQKnOVmgHt96HSf0aP37X1Dd8eWxPEzmcxed4RNkWdpXMeNOXd3YkyXpjg6VJxLzSX0hRB2V+LyxxaLMQkbtgIiVkN2CrjWhra3GT36gEE3fHdsSZzNzOHDjcf5bk8cNV2ceGlkGyb2DcDN2U6LsxRBQl8IYVc2lz/WGpL2WS+xXAmZSeBUw3p37FhoOQycy/dGx6zcfOZtj2b+jmjy8i081NufqUNaUt+j4i6NalPoK6VGAh8DjsAXWutZVz0/APgI6AiM01ovv+r52kAksEprPaU0Gi6EqNyuLn88q/+swssfn4n466aptBhwcLauHTvDuDvWtfRLFRQn32zhu9B4Ptx4nOQLudzSoTEvjmiDv1fp3KlblooNfaWUI/ApMBxIAPYopdZorSMKbBYHTABeuMZhZgDbbqypQoiqotjyxyknrEsKroBzkaAcIGAg9H8e2t4KNerZpd1aazZFnmXW2khOnMsi2K8e8x7qRldf+7TnetjS0+8BRGmtowGUUt8CdwCXQ19rfdL6nOXqnZVS3YCGwDqg2AL/Qoiqq8jyx+mJRtCHrYCk/cZjvr3h5jnQ7g7wsO+qcwfiz/Pur5HsjkmluVdNPn+wGze1a1jp6oHZEvpNgfgCPycAPW05uFLKAfg38CAwtIjtJgOTAXx9fW05tBCikrm6/PHzwc9TJ98Eu+cbY/Rxu4wNG3eG4TOMCdk69q9LFZeSzXvrj/DzoVN4ebgwY3R7xnVvhvMNLGRiT7aEfmFvY7bWA34S+FVrHV/Uu6HWeh4wD4zlEm08thCiEvhb+eMBH9Dn/Bn4foKxEIk2g1cbGPyaMSFbhnfHlkRaVh6fbI5i6R8ncXJw4KkhLZk8sAUerpX7+hdbWp8AFLxQ1gdIsvH4vYH+SqknAQ/ARSl1QWv9SsmaKYSojHYl7eKtXdbyx97defr8BdyXjQdzHtT1g75PQ4e7oEG7crlpyhY5JjOLdp3k0y1RZOXmc09wM54d3pqGtatGCXRbQn8P0EopFQAkAuOA8bYcXGt9/6XvlVITgGAJfCGqvvTcdObsns3q6J/wV24sPptGl5jlUKsxdH/M6NE3LZ+7Y21lsWhWH0hkzvqjJKXnMCSwAa+MCqR1w1r2blqpKjb0tdb5SqkpwHqMSzYXaq3DlVLTgVCt9RqlVHdgFVAPuE0p9ZbWOqhMWy6EqHjM+YSEfsLbR5eRZsnl0fQMnsi5gGvQXQXujq14Nyz9djyZd3+NJOJUBh2a1mHOPZ3o08LL3s0qE6qiLdcXHBysQ0ND7d0MIYStLBaI/4OUQ18zMymE9W5OtDGZmV63G+06PWwsROJYvuWDbRV5KoOZa4+w/dg5fOrV4MURbbitYxMcKlDZBFsppfZqrYu9QrJyz0gIIexDa+OyyrAV6PBV/GJOY3Z9T7LcnJnqM4xH+k3H2Q43TdnqVPpF/r3hGCv2JVDL1YnXbm7LQ338cHWqeJ9CSpuEvhDCdmcj/7o7NjWa086uzGjWgu3akY71g5je752SlT8uZxk5Jj7beoIFv8WgNTzWvzlPDmpBXXcXezet3EjoCyGKlhr9V72bs+GgHND+/VjebggfnNlJvs7npa4lKH9sB3n5Fr7+M5b/bI4iNSuPOzo34YWb2tDMsxQWUq9kJPSFEH+XkQThq4wefeJe47FmPWHU+8T7dufNQ3PZnbjB9vLHdqK1Zm3Yad5bd4STKdn0bl6ff97clg4+9l3IxJ4k9IUQhqxk69qxKyF2J6ChUUcYPh2CxmCu3ZSvIr/ik5DJtpU/trPQk6m882sk++PO07qhB19O6M6gNt4Vtr3lRUJfiOosJx2O/AKHl0P0Vuvdsa1h0KtGGQSvVgCcOH+CN9Y9xKFzxZQ/rgBOnLvAe+uOsD78DA1quTJ7bAfGdvXBqZKWTShtEvpCVDdmk7HCVNgKOL7BenesL/R9yrp2bPvLN02ZLCYWHl7I54c+L7r8cQVwLjOX/4Qc5+vdcbg5OfD88NZM6h+Au4vEXEHyryFEdaI1rJxsVLP0aAjBk4yg9wn+292xESkRvLHzDY6mHS28/HEFkZ2Xz4IdMXy27QQ5+RbG9/DlqaGt8K5VcRcysScJfSGqk72LjMAf+AoMfKnQu2Nzzbn878D/WBS+6O/ljysQs0WzfG88H2w8xpmMXEYENeSlkYG08K649wdUBBL6QlQXZ8Jh3SvQfDAMfLnQxcILLX/sWrGudNFas/XoOWaujeTYmQt08a3L3PFd6e7vae+mVQoS+kJUB3lZ8MMEcKsDd877W+AXLH/cuGZjPh/+OX2a9LFPW4twOCGdd3+N5PfoFPzqu/Pf+7syqn2jCjnHUFFJ6AtRHfz6EiQfh4dW/20FqivKHwfex9Ndn8bduWLdtBSfms2cDUf58UASnjVdePO2dozv6YeLk1yRU1IS+kJUdQe/gwPLYMCL0HzQ5YfTc9OZEzqH1VGr8a/tz+JRi+nSoIvdmlmY9GwTc7ccZ/GuWJSCJwe14IlBLajtVjELuFUGEvpCVGXJUfDzs+Dbx5i8tQqJC+GdP94hNSeVSe0n8Y/O/8DVseJc7ZKbb2bJrljmbokiI8fE2K4+PH9TaxrXqWHvplV6EvpCVFWmHGMc38kVxn4Bjk6YLCbe/fNdlh9bTpt6bZg7dC7t6rezd0svs1g0Px1K4v31R0lIu8iA1t68OiqQto1r27tpVYaEvhBV1YbX4MxhGP891GnK+ZzzPLftOfac3sPE9hOZ0mUKzg4VZ5hk14lkZv56hMOJ6bRrXJulkzrQv5W3vZtV5UjoC1EVRfwIe76A3lOg9Qii06OZGjKVU1mneLffu9zW4jZ7t/CyY2cymbX2CJuPnKVJHTc+uKcTozs3rZQLmVQGEvpCVDVpJ+HHqdC0Gwydxq6kXbyw9QWcHZ1ZOGIhnRt0tncLATiTkcOHG4/xfWg8NV2deGVUIBP6+OPmXDHLM1cVEvpCVCX5ebB8ovH9XQv5JmoFs3fPpnnd5swdMpcmHk3s2z7gQm4+87adYP6OGPItFib0CWDKkJZ41qw+C5nYk4S+EFXJ5umQuJf8uxYy6+gyvjv6HYN8BjFrwCxqOte0a9NMZgvf7onn403HSL6Qx60dG/PiiDb41bdvu6obCX0hqopjG2DXJ2R0fYjnk9bxx6k/eCToEZ7u+rRdV7TSWrMh4gyz1x0h+lwWPfw9mf9QIF1869mtTdWZhL4QVUFGEqx6nLhG7fg/cwwJZxKZ3mc6Y1qNsWuz9sWlMfPXSPacTKOFd03mPxTMsLYNpGyCHUnoC1HZmfNhxaPsdjTzbG1wyE1n/vD5BDcKtluTTiZn8d76I/x6+DReHq68M6Y99wY3k4VMKgAJfSEqu+3vsTz1AO94e+Pn3pBPhn5itzVrU7Py+E/Icb76MxYnBweeHtqKxwY0x8NVoqaikP8TQlRi5hNbmHN4Hsu86tO3aW/eH/A+tVxqlXs7ckxmFvwWw2dbT5CVl8+93X15dlgrGtR2K/e2iKJJ6AtRSV1IO8mLIf/Hb3Vq8UDre3m+5ys4OZTvn7TZolm1P5F/bzjKqfQchrVtwMsjA2nVsPzfeIRtJPSFqITi0+OYumYssS4O/KvdJO7p/ky5t2H7sXPMXHuEyFMZdPSpwwf3dKZ3i4q3nKK4koS+EJXM3jN7eXbDE5jNF/ms+T30LOfAj0jKYObaSHYcT6aZZw3+c18Xbu3QWMomVBIS+kJUIqujVvPWrjfxycthrkcn/AZOK7dzJ52/yJwNR1m1P5Habs68fktbHuzth6uTlE2oTCT0hagEzBYzH+//mC/DvqSnSfPvLGfqPDAPyuF694wcE//dcoKFO2MAmNy/OU8Oakkd94pToVPYTkJfiAou25TNyzteZmv8Vu51qM/LSWE4T1pvrHdbhvLyLSz7I5ZPNh8nLdvEmC5Nef6m1vjUq1hLKYqSkdAXogI7deEUUzZPIep8FK82HMD4P5bBiHeNCpplRGvNL4dP8d66o8SlZtO3ZX1eHdWW9k3L9k1GlA+bbo9TSo1USh3PtTiQAAAgAElEQVRVSkUppV4p5PkBSql9Sql8pdRdBR7vrJT6XSkVrpQ6pJS6tzQbL0RVduDsAcb9Mo5TF07xvy4vMH7P99B6JPR6sszOuTsmldH/3cWUr/dTw9mRRY90Z9mknhL4VUixPX2llCPwKTAcSAD2KKXWaK0jCmwWB0wAXrhq92zgIa31caVUE2CvUmq91vp8qbReiCrq5+ifmbZzGg1rNmTu4Fk0//ZhqOkNo/9XJuP4UWcvMHvdETZGnKFhbVfeG9uRsd18cJQrcqocW4Z3egBRWutoAKXUt8AdwOXQ11qftD5nKbij1vpYge+TlFJnAW9AQl+IQli0hbn75zL/8HyCGwbz4cAPqPvLC8bCKBN+AXfPUj3f2cwcPt50nG/3xFPD2ZEXR7RhYt8AarjIFTlVlS2h3xSIL/BzAtCzpCdSSvUAXIAThTw3GZgM4OvrW9JDC1ElZJuyeX3n62yM3cidre7k9Z6v43zwGwhbDkNeB78+pXaurNx8vtgRw+fbT5CXb+H+nr48NbQVXh6upXYOUTHZEvqFfb7TJTmJUqoxsBR4WGttufp5rfU8YB5AcHBwiY4tRFVwJusMUzdP5UjqEV4MfpEH2z2IOncEfn0JAgZCv+dK7Vwbwk/z2uowzmXmMqp9I14c0Ybm3h6ldnxRsdkS+glAwZJ9PkCSrSdQStUGfgFe11r/UbLmCVH1hSeHM3XzVLJMWcwdOpcBPgMgLxt+mACuHnDnfCiFRVAsFs1/Nh/no03Had+0Np890JVufqU7XCQqPltCfw/QSikVACQC44DxthxcKeUCrAKWaK1/uO5WClFFrTu5jtd/ex2vGl58PvxzWtVrZTyx9iU4dxQeXAm1Gt7weS7k5vP89wdYH36GsV19eGdMe1mAvJoqNvS11vlKqSnAesARWKi1DldKTQdCtdZrlFLdMcK9HnCbUuotrXUQcA8wAKivlJpgPeQErfWBsngxQlQWWms+O/QZ/z3wX7o06MKHgz6kfg1rsbJDP8D+pdD/eWgx5IbPFZeSzWNLQjl+NpN/3dqOiX39ZeWqakxpXbGG0IODg3VoaKi9myFEmcnJz+GNnW+w9uRabm9xO9N6T8PF0cV4MuUEfD4AGrY3rtZxvLH7J387nsyUb/ahNXw6viv9WnmVwisQFZFSaq/Wutjl0uSOXCHK0bnsczy1+SnCU8J5puszTGw/8a9ed36uMY7v6Ax3LbihwNdas3DnSd75JYKWDTyY/1AwfvVrls6LEJWahL4Q5SQyJZKpm6eSkZfBh4M/ZKjv0Cs32PAvOH0I7vsW6vhc93lyTGZeWxXGin0J3NSuIR/c21mWKxSXyW+CEOUgJC6EV3e8Sm2X2iwZtYRAz8ArN4j8CXZ/bpRYaDPqus9zJiOHyUv3cjD+PM8Ma8VTQ1pJnXtxBQl9IcqQ1poFYQv4eN/HdPTqyMdDPsarxlXj6ufj4Mf/gyZdYNhb132ufXFpPLF0Lxdy8/nsgW6MbN/oBlsvqiIJfSHKSJ45jzd3vclP0T8xKmAU0/tMx83pqoXCzSZYPhEsFrhrITi5XNe5vg+N5/VVYTSq48aSST0IbFS7FF6BqIok9IUoAykXU3hmyzMcOHeAKZ2nMLnj5MIvk9z8NiTsgbu+BM/mJT5PvtnCO79G8uXOk/Rr6cXc8V2o6359bxyiepDQF6KUHUs7xtSQqaTmpDJn4BxG+I8ofMPjm2DnR9BtArS/s8TnScvK4/++3seuEylM7BvAP28OxMnRpmrpohqT0BeiFG2L38ZL21/Cw9mDRSMXEeQVVPiGGadg1ePQoB2MnFXi8xw5ncFjS0I5k5HLnLs7cVe367/aR1QvEvpClAKtNUsilvDv0H8T6BnIJ0M+oWHNa5RPsJhh5WNgyoa7F4FzjRKda+3hUzz/w0E8XJ34bnIvuvjWu/EXIKoNCX0hbpDJbGLGHzNYFbWK4X7DeaffO9RwKiLIt8+Bkzvgjv+Cdxubz2OxaD4KOc5/Qo7TuVldPn+wGw1ruxW/oxAFSOgLcQPSctJ4duuz7D2zl8c7Ps6TnZ/EQRUxrh6zA7bNgo73Qmeb6hYCRsG0Z787wMaIM9zdzYcZo6Vgmrg+EvpCXKcT508wJWQKZ7PPMqv/LG5pfkvRO2Qlw4pHjat0bvm3zcsenkzO4rEloUQnZ/Hmbe14uI8UTBPXT0JfiOvwW+JvvLjtRVwdXVk4ciGdvDsVvYPFAquegItpcP8P4FrLpvNsP3aOKV/vw8FBsXRiD/q0lIJp4sZI6AtRAlprvj7yNe/teY9WdVvxyZBPaOzRuPgdf/8EojbCzXOgcUebzrPgtxje/TWS1g1rMe/BYHzru5fCKxDVnYS+EDYyWUzM+nMW3x/7nkHNBjG7/2zcnW0I4vg9EDId2t4O3R8tdvMck5l/rjzMyv2JjGrfiDl3d6KmFEwTpUR+k4Qohtaa30/9zn8P/JeD5w4ysf1Enu76dNETtpdcTDPKLNRuArd/Uuw4/un0HB5fGsrBhHSeG96aKYNbSsE0Uaok9IW4BrPFzMbYjSwMW0hkaiReNbx4t9+73NbiNtsOoDWsmQqZSTBxPdSoW+Tme2PTeGLZXrJz85n3YDduCpKCaaL0SegLcZWc/BzWnFjDovBFxGfG41/bnzd7v8ltLW77a4UrW+z5wiiZPHwG+BS9oNF3e+L41+pwGtd146tHe9K6oW0TvUKUlIS+EFbpuel8f/R7lkUuIzUnlQ5eHXiu23MMbjYYR4cSXhN/6hCs/ye0ugl6T7nmZiazhbd/jmDx77H0b+XFJ/dJwTRRtiT0RbV3JusMSyOW8sOxH8jOz6Zv075MDJpI90bdr+96+NxMY9lD9/ow+jNwKHzsPzUrjye/2ssf0ak81j+Al0dKwTRR9iT0RbUVfT6aL8O/5Ofon7FoCyP8RzCx/cS/r2pVElrDz89BWgw8/BPUrF/oZhFJGUxeGsrZzFw+uKcTd3aVgmmifEjoi2rnwNkDLAxbyJb4Lbg5unFXq7t4OOhhfGqVQvAe+AoOfw+D/gn+/Qrd5JdDp3jhh4PUqeHMD4/3plOzoid4hShNEvqiWtBasyNxBwsOL2Df2X3UdqnN4x0fZ3zb8Xi6eZbOSc4egV9fBP/+MOCFvz1tsWg+3HSMTzZH0dW3Lp892I0GtaRgmihfEvqiSjNZTKyLWcfCsIVEnY+iUc1GvNT9Jca2GmvbjVW2ysuG5Y+AszuM/QKumvjNzDHx7HcH2BR5lnuDmzF9dBCuTlIwTZQ/CX1RJWWbsll5fCVLIpZwKusULeu25J1+7zAqYBTODs6lf8J1r8DZCHhgBdS68vr6GGvBtJjkLKbfEcSDvfykYJqwGwl9UaWk5aTxzZFv+PrI16TnptO1QVde6/ka/X3623YH7fU4vBz2LYZ+z0LLYVc8te3YOaZ+vQ9HB8WyST3p3aLwiV0hyouEvqgSki4ksTh8MSuPryTHnMOgZoOY1H4SnRt0LtsTp5yAn54Bnx4w+LXLD2utmb8jmllrj9C6YS3mPxRMM08pmCbsT0JfVGpHU4/yZfiXrItZh0JxS/NbeKT9I7So26LsT56fa9TVcXCEuxaAozFslGMy88qKQ6w+kMQtHRrz/t0dcXeRPzVRMchvoqh0tNaEngllYdhCfkv8jRpONbi/7f082O5BGtUsx3o1G6fBqQNw71dQ1xeApPMXeXzpXsKS0nlxRBueHNRCxu9FhSKhLyoNi7awJW4LC8MWcij5EJ5unkzpPIVxgeOo41qnfBtz5Ff483/Q8wloeysAoSdTeWLZXnJMFuY/GMywdtdYGF0IO5LQFxVenjmPn6N/5suwLzmZcZKmHk15redrjG45GjcnO1znfj4eVv8DGneC4dMB+GZ3HG/8GIZPPXe+ndyNlg2kYJqomCT0RYV1Ie8CPxz7gWURyzh78SyBnoG8N+A9hvsNx8nBTr+6ZhOsmAQWM9z1JSblzPTVYSz9I5YBrb35ZFwX6riXwSWhQpQSm/5ylFIjgY8BR+ALrfWsq54fAHwEdATGaa2XF3juYeB1649va60Xl0bDRdWVfDGZZRHL+P7o92SaMunZqCcz+s6gd5Pe9h8f3/IuxP8JYxeQ4urDP774k90xqTw+oDkvjQzEURY8ERVcsaGvlHIEPgWGAwnAHqXUGq11RIHN4oAJwAtX7esJTAOCAQ3ste6bVjrNF1VJXEYci8IX8WPUj5gsJob5DWNi+4m092pv76YZokLgtw+h60OE1x/O5Lk7Sb6Qy0f3dmZ0l6b2bp0QNrGlp98DiNJaRwMopb4F7gAuh77W+qT1OctV+44ANmqtU63PbwRGAt/ccMtFlRGeEs7CwwvZGLsRJwcn7mh5BxOCJuBX28/eTQOLBWK2wt7FcOQX8A7k16bP8Nz/dlHP3YXlT/Shg085TyILcQNsCf2mQHyBnxOAnjYev7B9/9YlUkpNBiYD+Pr62nhoUZldWnd2YdhC/jz1Jx7OHkxsP5EH2j2AVw0vezcPMk8bFTP3LobzsVCjHpbuj/GZaRTv/XCEYL96/O+BbnjXcrV3S4UoEVtCv7BBSm3j8W3aV2s9D5gHEBwcbOuxRSWUb8lnU+ymy+vOetfw5tluz3J367up5WLnK14sZjixGfYugqNrQZvBvz96yL8Ide/Hh1ti2XUihft6+PLW7UG4OMmCJ6LysSX0E4BmBX72AZJsPH4CMOiqfbfauK+oQnLyc/gx6kcWhS8i4ULC9a87WxYykmD/Mti3BNLjwd0Lev8f+Z0fZP3pWszbEc3B+H141nTh3TEdGN9TPo2KysuW0N8DtFJKBQCJwDhgvI3HXw+8q5SqZ/35JuDVErdSVFrpuel8d/Q7vor86vK6s88HP399686WJnM+RG00hm+OrwdtgeaD4KYZZAWM4If9Z1iwKIb41Iv413fn7dHtGdvVhxouUg5ZVG7Fhr7WOl8pNQUjwB2BhVrrcKXUdCBUa71GKdUdWAXUA25TSr2ltQ7SWqcqpWZgvHEATL80qSuqttNZp1kasZTlx5ZfXnd2UvtJBDcMtu9ll+fjYN9So2efmQQeDaHvM9D1Qc46N2HxrpMsW/4b6RdNBPvV4/Vb2jGsbUO5FFNUGUrrijWEHhwcrENDQ+3dDHGdCq47q7W+vO5sG8829muU2QTH1hm9+qhNxmMth0G3h6H1SI4n5/DFjhhW7U/EZLEwol0jHhvQnG5+9Yo+rhAViFJqr9Y6uLjt5I5cUSquXnf27tZ383DQwzT1sOP166kxxjj9ga/gwhmo1QQGvgRdHkDXacYf0anMX3qAzUfO4ubswL3dmzGpXwD+XjXt12YhypiEvrhuV687W8e1Dk90eoL7Au8rvXVnSyo/D47+YvTqo7eAcoBWI4xefcvh5OPAr2Gnmb99J4cT06lf04XnhrfmgV5+eNa084SyEOVAQl+UWGHrzr7c/WXubHVn6a47WxIpJ4xLLQ98DdnJUKcZDPondHkA6jTlQm4+3/0ez8LfYkg8f5Hm3jWZeWcHxnRpipuzTM6K6kNCX9issHVn3+33LiMDRpbNurPFMeXAkZ+NsD+5A5QjtBkF3SZAiyHg4MiZjBy+XHuEr/6MJTMnnx4Bnrx1exBDAhvgIJOzohqS0BfFSstJ4+sjX/PNkW8urzv7eq/X6d+0v32uxDl31Bi+Ofg1XEyDun4w9A3ofP/lRcmPnM5g/vYY1hxMxGzRjOrQmMf6N6dzs7rl314hKhAJfXFNiRcSWRy+mFXHV5XvurOFMV2EiB+NXn3c7+DgDIG3GL36gIHg4IDWml1RyXy+PZrtx85Rw9mR+3v6MbFvAL71ZX1aIUBCXxTiaOpRFoYtZP3J9SiluCWgHNedvdqZcKNXf+hbyEkHz+bGwiWdxoOHNwAms4Vf9icyb3s0Eacy8PJw5cURbbi/py913WVyVoiCJPQF8Pd1Z92d3O2z7ixAXhaErzJ69Ql7wNEF2t5u9Or9+4F1SCkzx8S3u+NZuDOGU+k5tGzgwXtjO3JHlya4OsnkrBCFkdCv5gpbd3Zql6nc2+be8l939tQhI+gP/wC5GeDVGka8Cx3HQc36lzdLOn+RRbtO8s2fcWTm5tO7eX3eHdOBga29ZXJWiGJI6FdTV6876+Phw+s9X+eOlneU77qzuZkQtsII+6T94OQG7UYbvXrfXpd79QDhSel8sSOGnw4moYFbrJOzUs9eCNtJ6FczV68729azLe8PeJ9hfsPKb91ZrSFpnzFWf3g5mLKgQTsY9R50vAdq1CuwqWbH8WTmbY/mt6hk3F0ceai3PxP7+eNTTyZnhSgpCf1q4m/rzjbuyYx+M+jduBzXnc1JN4Zu9i6C04fB2R2C7jR69T7BV/Tq8/It/HQwifk7ojlyOpMGtVx5eWQg43v4ysLjQtwACf0qLjYjlkXhi1gTtQaTxcRwv+FMbD+RIK+g8mmA1sZk7N7FEL4STNnQqAPc8m/ocDe4XTk0k37RxDe74/hyZwxnMnJp07AWc+7uxO2dmsiiJUKUAgn9Kio8OZwFYQvYFLsJZwfny+vO+tYupwVALqbBoe+NXv3ZCHDxMEK+2wRo0uWKXj1AQlo2X+48ybe748jKM9OvpRfv3dWJAa287FuKWYgqRkK/CtFa83uSdd3Z039Sy7kWkzpM4v6295fPurNaGzdO7V0MEashPweadIXbPob2Y8H178shhiWmM297NL8cPoUCbuvUhEf7BxDURCZnhSgLEvpVQGHrzj7X7Tnubn03Hi4eZd+ArBQ4+A3sWwzJx8C1tlHorOvD0Ljj3zbXWrP16DnmbY/m9+gUPFydmNjXn0f6BtCkbo2yb68Q1ZiEfiVW2Lqzb/V5i1ub31r2685qbRQ527sIIn8Ccx749IA7PoWgMeDy95r0uflmfjyQxPzt0Rw/e4FGtd34582BjOvhS203mZwVojxI6FdCV68729GrIy8Ev8Bg38E4qDKe7LxwzliUZN8SSD1hTMQGTzR69Q3bFd7ebBPL/oxl0a6TnMvMpW3j2nx4bydu6SCTs0KUNwn9SuTqdWf7Ne3HxPYTy37dWYsFYrYavfojv4LFBL59jFWo2t0BzoUPycSnZrPgtxi+D40nO8/MgNbefHhPc/q2rC+Ts0LYiYR+JRB9PpqFYQv5JeYXtNaMDBjJI0GPlP26s5mnjQXE9y2B87FQwxN6Pg5dHwLva5/7YPx55u2IZu3hUzg6KG7v1JRH+wfQtnHtsm2vEKJYEvoV2IGzB1gQtoCt8Vtxc3Tjntb38FDQQ2W77qzFDCc2G736o2tBm8G/v1Gvvu1t4ORa+G4WzeYjZ5m3I5rdManUcnNi8oAWTOjjT6M65VjWQQhRJAn9CsaiLexI2MHCsIWX1539R6d/cF/gfdRzq1f8Aa5XeqLRq9+/FNLjwd0L+kwxxurrX7ukco7JzKr9iczfEU30uSya1q3B67e0ZVwPXzxc5ddLiIpG/iorCJPFxNqYtXwZ9iVR56NoXLNx2a87a86HqI1Gr/74BtAWaD4Ybnob2twMTte+AigtK4+lf8Sy5PeTJF/Io33T2nw8rjM3d2iMs6NMzgpRUUno21m2KZsVx1ewJGIJp7NOl8+6s+fjYN9So1efeQo8GkK/Z6HLg+AZUOSusSlZlydnc0wWBrfx5rEBzendXCZnhagMJPTtpLB1Z//V619lt+6s2QTH1hm9+qgQ47GWw+DmOdB6BDgW/QazLy6N+dujWRd+GmcHB0Z3acKj/ZvTuuHf77IVQlRcEvrl7Op1Zwc3G8zE9hPLbt3Z1Bjj6psDX8GFM1CriXGpZZcHoG7RdXjMFs2myDPM3x5NaGwatd2c+MdAY3K2QW2ZnBWiMpLQLydXrzt7a/NbeSToEZrXbV76J8vPgyM/G2URoreCcoBWI4xiZy2HgWPR/9tzTGaW701gwW8xxCRn4VOvBtNua8c9wc2oKZOzQlRq8hdchi6tO7sgbAE7E3fi7uTOA20f4IF2D5TNurPJUUbQH/gaspOhTjMY/Bp0vh/qFH+ZZ8qFXJb8HsvSP2JJzcqjo08d5o7vwsigRjjJ5KwQVYKEfhnINmWzLWEbSyOWcjj5MJ5unjzV5SnuaXNP6a87a8oxat/sW2zUwlGO0GYUdHsEWgwGh+IXCI8+d4EFv8WwfG8CufkWhrVtwGP9m9MjwFMmZ4WoYiT0S0l6bjrbEraxKXYTu5J2kWvOxcfDh3/1+he3t7i99NedPXfUKGF88Gujdn09f+MGqs73Q63iP0Vordkbm8a87dFsjDyDs6MDY7s2ZVK/5rRsUA6VOYUQdiGhfwOSLyazOW4zm2I3sef0HvJ1Pg3dG3JX67sY6juUrg264mhDT9tmposQvtro1cf9Dg7O0PZW4waqgIHgUPwQjNmi2RB+mnk7otkfd5667s5MHdySB3v7412r8LtthRBVh4R+CSVkJhASF0JIXAgHzh5Ao/Gr7cfDQQ8zzG8YQfWDSn9I5Ey4canloe+MdWY9W8DwGdDpPvDwtukQ2Xn5LN+bwBc7YohLzcavvjsz7ghibDcf3F3k10CI6sKmv3al1EjgY8AR+EJrPeuq512BJUA3IAW4V2t9UinlDHwBdLWea4nWemYptr/Maa2JTo9mU+wmQuJCiEyNBCDQM5AnOz/JMN9htKjbovSDPi8LwlYaYZ8YCo4uRkXLrg+Df7+/LTd4Lecyc1ny+0mW/hHL+WwTnZvV5dVRgdwU1AhHBxmvF6K6KTb0lVKOwKfAcCAB2KOUWqO1jiiw2SQgTWvdUik1DpgN3AvcDbhqrTsopdyBCKXUN1rrk6X9QkqT1pqIlAg2xW1iU+wmTmacBKCzd2deCH6BIb5DaFarWdmc/NRBa6/+B8jLBK82MGImdBoH7p42HybqbCZf7Ihh5f5ETGYLw9s2ZPKA5nTzqyeTs0JUY7b09HsAUVrraACl1LfAHUDB0L8DeNP6/XJgrjKSRQM1lVJOQA0gD8gonaaXLrPFzP6z+y8P3ZzKOoWjcqR7o+480PYBBvsOpoF7g7I5eW4mHF5uhP2pA+DkZqw+1fVh8O1lc69ea83umFTmbY8m5MhZXJ0cuLubD5P6BdDcWyZnhRC2hX5TIL7AzwlAz2tto7XOV0qlA/Ux3gDuAE4B7sCzWuvUq0+glJoMTAbw9S36LtHSZDKb+PP0n2yK3cSW+C2k5qTi4uBCn6Z9+L/O/8dAn4HUdatbNifXGpL2GUF/eAWYsqBBEIx6HzreDTVsr6iZb7awLvw087dHczAhHc+aLjwzrBUP9vKjvodMzgoh/mJL6BfWzdQ2btMDMANNgHrADqXUpkufGi5vqPU8YB5AcHDw1ccuVdmmbHYl7WJT3Ca2x28n05SJu5M7A30GMtRvKP2a9qOm89/Xdy01Oelw6Hvjcsszh8HZHdrfaVxX37Sbzb16gKzcfL4PjWfBbzEkpF0kwKsmb49uz13dfHBzLsWrhoQQVYYtoZ8AFBzA9gGSrrFNgnUopw6QCowH1mmtTcBZpdROIBiIphxl5GWwLX4bIXEh7EzcSY45h7qudRnmN4xhfsPo2bgnro5l2CPWGhL2GL36sJWQfxEadYRbPoAOd4NbyVaUOpuRw6JdJ1n2RywZOfkE+9XjX7e2Y1jbhjI5K4Qoki2hvwdopZQKABKBcRhhXtAa4GHgd+AuYLPWWiul4oAhSqllGMM7vYCPSqvxRUm+mMyW+C2ExIbw56k/ydf5NKjRgDGtxjDMdxhdG3bFyaGML1XMTrX26hfBuUhw8YBO9xo1cJp0KfHhjp3JZP72aH48kITJYmFkUCMe7W9MzgohhC2KTT3rGP0UYD3GJZsLtdbhSqnpQKjWeg2wAFiqlIrC6OGPs+7+KfAlEIYxBPSl1vpQGbwOAJIuJBESF8Km2E3sP7sfjaZZrWY8GPQgw3yH0d6rPQ6qjGrIWMyQdtK4pv5sJJw+BMc3gjnXGLa57T/Qfiy4lmxCVWvN7ydSmLcjmq1Hz+Hm7MC4Hs2Y2DcAf68yHIYSQlRJSusyHUIvseDgYB0aGmrz9tHp0YTEhrApbhMRKcYFRa3rtWaY7zCG+g2lVd1WpXuJotbGwiNnI+BMhBHwZyOMsgj5F60bKaMsQsuhxhU4jTuW+DQms4VfD59i/o5owhIz8PJw4eHe/jzQy496Na+9opUQonpSSu3VWgcXt12luxVTa01kauTlm6Wi043pgY7eHXmu23MM9R2Kb+1SugLoYtpfoX420hryEZBz/q9tPBpBg7bQfZLx3wbtwLsNuFxfL/xCbj7f7o7jy50nSTx/kebeNZl1ZwdGd2kqk7NCiBtWKULfbDFz8NxBNsVtIiQ2hKSsJByVI8ENgxkXOI4hzYbQsGbD6z+B6aLRUz8bCWfD/wr4zALz1a61jUAPGgMNg/4K+BLcMFWU0+k5fLkrhq//jCMzJ58eAZ68dXsQQwIb4CCTs0KIUlJhQ99kMbHn1B42xW1ic9xmUnJScHZwpk+TPjzR6QkGNRtEPbcSTmCa8yEt5q9x90sBnxptLAoO4OgK3q0hoL8R6g3aQcN2ULtpiS6ntFXkqQzm74hmzYEkLFozqkNjHuvfnM7Nyuj+ACFEtVbhQj8zL5N/7vgnWxO2kpmXSQ2nGgzwGcAw32H0a9oPDxcbJkK1hoxEa4+9QMCfO2ZMrAKgwLO5Eejtx/4V8J7Ni11Z6kZprfktKpl526PZcTwZdxdHHujlx6R+ATTzdC/TcwshqrcKF/pxmXFsS9jGkGZDGOY3jF6NexVdiz47tcCY+6WAj4Tc9L+2qdXEGI5pPuivcPduA841yvrlXMFktvDzoSTmbY8h8lQG3rVceXHE/7d3L6FxVXEcx7//NtOZtpnpM5PaxtIHNaYlPhAsVvABRSouWkVoxYVKKXShewX34sZFQZGKtbsWdeFjIW7FUqhFofiEapz/E0EAAATeSURBVKUNtZnUtE5qMjOd5LiYSTJJKvO8c2/m/D4Qktw5d+65c//5cTh3MqefF3dvZvUK3ZwVkeBF7t07A/cPuAs/XCC2JDb3gcI4jPw6G/DT7565dW22TWJV6aMM0gOlr95d0HNvy+bdG5XN3eb0ucuc+PZPrmVz7Eh3c+Sxbex/YCPxLt2cFZHmLdp376yMrSB2/fe54Z75GUYvMfPpD12J0kh9+5Nz592TdwUy796oqzcn+OjMJU6du8KtfJFHtq3jrecGefyeHt2cFZFQRC70+esCvFf+PDdbUlowZMMg3HewYt59a01rvwZtasoxOl5gOJsjk82TGcsxnM0znM0xdGOCMxev44BnyjdnB/tavD6uiEidohf6K3vg2XdK0zPr+yHW4rVla+Cc48b47TkhnslW/DyWJ1P+XpxaOD22duUy0sk4L+3ZwiuPbqFvjW7Oikg0RC/0UxtLC4YEwDlHdqLI8FiO4ezcQM+M5We2jYzlKUxOLdh/9YoY6WSc3lSC7T3r6U2Vfu5NxelJTn+Pa55eRCIreqHfAOccY/nigtH49LTLcDZXDvo8heLCME8mumbCe/fWtaRTiZlwnw72nmRc/xErIote5EP/Vr5YHoHnGKkYjc+fR5+4Pblg3+54F+lUnN5kgoc2r5kX5qVATycTLF+mMBcRP0Qu9K+MjnPo+NmZEfq/hYVhvjy2lA2rSqPvwb7V7C0Hebo8Kk8n46RTCbrjkTs9EZFQRS4VxwuTTE45BjameKI/XRqNl0fr6fLovDvepcW9RUQaELnQ79+Q5JOje8LuhohIRwpoRREREYkihb6IiEcU+iIiHlHoi4h4RKEvIuIRhb6IiEcU+iIiHlHoi4h4JHIrZ5nZBPBTA7uuAv6p2qq+ttXaVXt8PXC9xj5FWT2vbdSP2+xzbgYuB3jMVtVmtTadUpsQTn1GsTZ3OeeqrwHrnIvUFzDS4H7HW922WrsaHj8f9uvZomtS82sb9eM2+5yN1GcYtVmtTafUZiuuaVSO2a7ajOL0zs0G9/sygLbV2tVzzMUsrPMM4rjNPmcj9RlGbdZ73MUsjPNctLUZxemd866GxX0Xg046FynplGvaKechs2q9plEc6R8PuwMt1EnnIiWdck075TxkVk3XNHIjfRERCU4UR/oiIhIQhb6IiEcU+iIiHlHot5GZbTOzD83s07D7IlLJzA6Y2Qdm9rmZPRV2fyQ4Cv0mmdkJM8uY2Y/ztu8zs9/M7KKZvQ7gnPvDOXc4nJ6Kb+qszc+cc0eAl4GDIXRX2kSh37yTwL7KDWa2FHgXeBrYCbxgZjvb3zXx3Enqr803y49Lh1LoN8k59w0wOm/zw8DF8si+AJwG9re9c+K1emrTSt4GvnLOfd/uvkr7KPSDsQm4UvH7ELDJzNaZ2fvAg2b2RjhdE8/dsTaB14C9wPNmdjSMjkl7dIXdgQ5ld9jmnHN/A/qDkjD9X20eA461uzPSfhrpB2MIuLvi9z7gakh9Eamk2vScQj8Y3wE7zGyrmS0DDgFfhNwnEVBtek+h3yQzOwWcBfrNbMjMDjvnisCrwNfAL8DHzrlGFoYRaZhqU+5EH7gmIuIRjfRFRDyi0BcR8YhCX0TEIwp9ERGPKPRFRDyi0BcR8YhCX0TEIwp9ERGPKPRFRDzyH/32BV/BgLSAAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "for col in eval_measure:\n",
    "    df_time = plot_metric(col)\n",
    "print (df_time)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "save_pickle(os.path.join(PARAM_DIR, model_type), 500, False)\n"
   ]
  },
  {
   "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}