notebook2.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Mol2Vec Experiments\n",
    "***"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:00.830936Z",
     "start_time": "2019-08-13T23:12:58.499879Z"
    }
   },
   "outputs": [],
   "source": [
    "from rdkit import Chem\n",
    "from mol2vec.features import mol2alt_sentence, MolSentence, DfVec, sentences2vec\n",
    "from mol2vec.helpers import depict_identifier, mol_to_svg, IdentifierTable, plot_2D_vectors\n",
    "\n",
    "from gensim.models import Word2Vec\n",
    "\n",
    "model = Word2Vec.load('db/model_300dim.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:01.724945Z",
     "start_time": "2019-08-13T23:13:01.719889Z"
    }
   },
   "outputs": [],
   "source": [
    "suppl = Chem.SDMolSupplier('db/chembl_01.sdf')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now, let's create a function that converts the suppl into a list of molecules (Mol type)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:02.284251Z",
     "start_time": "2019-08-13T23:13:02.278525Z"
    }
   },
   "outputs": [],
   "source": [
    "def suppl_to_lst(suppl, sample_size):\n",
    "    # declare a list to store the mols\n",
    "    ms = []\n",
    "    # sample size: can be changed accordingly to the size of database\n",
    "    for i in range(sample_size):\n",
    "        ms.append(suppl[i])\n",
    "    \n",
    "    return ms"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create different sample sizes from suppl"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:05.266368Z",
     "start_time": "2019-08-13T23:13:03.208902Z"
    }
   },
   "outputs": [],
   "source": [
    "small_moldb100 = suppl_to_lst(suppl, 100)\n",
    "small_moldb200 = suppl_to_lst(suppl, 200)\n",
    "small_moldb500 = suppl_to_lst(suppl, 500)\n",
    "small_moldb1000 = suppl_to_lst(suppl, 1000)\n",
    "small_moldb2000 = suppl_to_lst(suppl, 2000)\n",
    "small_moldb5000 = suppl_to_lst(suppl, 5000)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Clean the database"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:09.932278Z",
     "start_time": "2019-08-13T23:13:09.919774Z"
    }
   },
   "outputs": [],
   "source": [
    "def clean_db(db, radius):\n",
    "    not_in_model = []\n",
    "    none_type = []\n",
    "    \n",
    "    # check for nonetype\n",
    "    for i in range(len(db)):\n",
    "        mol = db[i]\n",
    "        # check for nonetype\n",
    "        if type(mol) == type(None):\n",
    "            none_type.append(mol)\n",
    "        \n",
    "    cleaned1 = list(set(db)-set(none_type))\n",
    "    \n",
    "    for i in range(len(cleaned1)):\n",
    "        mol = cleaned1[i]\n",
    "        # check if sentence in model or not\n",
    "        sentences = mol2alt_sentence(mol, radius)\n",
    "        for s in sentences:\n",
    "            if s not in model:\n",
    "                not_in_model.append(mol)\n",
    "    \n",
    "    cleaned = list(set(cleaned1)-set(not_in_model))\n",
    "    return cleaned"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Clean the databases, with radius = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:13.658421Z",
     "start_time": "2019-08-13T23:13:10.780017Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/sabrinaho/anaconda3/lib/python3.7/site-packages/ipykernel_launcher.py:19: DeprecationWarning: Call to deprecated `__contains__` (Method will be removed in 4.0.0, use self.wv.__contains__() instead).\n"
     ]
    }
   ],
   "source": [
    "radius = 1\n",
    "c_small_moldb100 = clean_db(small_moldb100, radius)\n",
    "c_small_moldb200 = clean_db(small_moldb200, radius)\n",
    "c_small_moldb500 = clean_db(small_moldb500, radius)\n",
    "c_small_moldb1000 = clean_db(small_moldb1000, radius)\n",
    "c_small_moldb2000 = clean_db(small_moldb2000, radius)\n",
    "c_small_moldb5000 = clean_db(small_moldb5000, radius)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "***"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:20.716560Z",
     "start_time": "2019-08-13T23:13:20.701966Z"
    }
   },
   "outputs": [],
   "source": [
    "from IPython.display import SVG\n",
    "from rdkit.Chem import rdDepictor as rdd\n",
    "from rdkit.Chem.Draw import rdMolDraw2D as draw2d\n",
    "\n",
    "def draw2D(x):\n",
    "    m = x\n",
    "    molSize = (450, 150)\n",
    "    mc = Chem.Mol(m.ToBinary())\n",
    "    if not mc.GetNumConformers():\n",
    "        rdd.Compute2DCoords(mc)\n",
    "    drawer = draw2d.MolDraw2DSVG(molSize[0],molSize[1])\n",
    "    drawer.DrawMolecule(mc)\n",
    "    drawer.FinishDrawing()\n",
    "    svg = drawer.GetDrawingText()\n",
    "    display(SVG(svg.replace('svg:','')))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:21.468045Z",
     "start_time": "2019-08-13T23:13:21.449067Z"
    }
   },
   "outputs": [],
   "source": [
    "from rdkit import DataStructs\n",
    "from rdkit.Chem.Fingerprints.FingerprintMols import FingerprintMol"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "After getting a list of molecules, let's create a function to find the Tanimoto similarity:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:23.497383Z",
     "start_time": "2019-08-13T23:13:23.489209Z"
    }
   },
   "outputs": [],
   "source": [
    "def tanimoto_similarity(ms, idx):\n",
    "    # convert Mol to datastructure type using a list comprehension\n",
    "    fps = [FingerprintMol(i) for i in ms]\n",
    "\n",
    "    # set a query molecule: can be changed \n",
    "    query = fps[idx]\n",
    "    # declare new list for similarities\n",
    "    sim = []\n",
    "    # loop through to find Tanimoto similarity\n",
    "    for i in range(len(fps)):\n",
    "        # tuple: (idx, similarity)\n",
    "        sim.append((i, DataStructs.FingerprintSimilarity(query, fps[i])))\n",
    "\n",
    "    # sort sim using the similarities\n",
    "    sim.sort(key=lambda x:x[1], reverse=True)\n",
    "\n",
    "    return sim"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Before getting into embedding molecules into vectors, I'm going to breifly introduce Word2Vec, which Mol2Vec is based on and is what I will be using."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## What is [Word2Vec](https://en.wikipedia.org/wiki/Word2vec)?\n",
    "Word2Vec is a two-layer neural networks that input a word corpus and output a vector space. Although Word2Vec is not a deep neural network, it converts words into numerical format that deep nets can understand.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## [Mol2vec](https://github.com/samoturk/mol2vec)\n",
    "The concept of Mol2Vec is based from Word2Vec, instead of words, it's molecules. \n",
    "\n",
    "Now, let's load a molecule database from [chEMBL](ftp://ftp.ebi.ac.uk/pub/databases/chembl/ChEMBLdb/) (same as the one mentioned above)\n",
    "#### Load molecule database\n",
    "- len(suppl) = 440055"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:25.574569Z",
     "start_time": "2019-08-13T23:13:25.569196Z"
    }
   },
   "outputs": [],
   "source": [
    "suppl = Chem.SDMolSupplier('db/chembl_01.sdf')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now, let's load the first mol in suppl and display it:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:26.726810Z",
     "start_time": "2019-08-13T23:13:26.708649Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/svg+xml": [
       "<svg baseProfile=\"full\" height=\"150px\" version=\"1.1\" width=\"450px\" xml:space=\"preserve\" xmlns=\"http://www.w3.org/2000/svg\" xmlns:rdkit=\"http://www.rdkit.org/xml\" xmlns:xlink=\"http://www.w3.org/1999/xlink\">\n",
       "<rect height=\"150\" style=\"opacity:1.0;fill:#FFFFFF;stroke:none\" width=\"450\" x=\"0\" y=\"0\"> </rect>\n",
       "<path d=\"M 246.47,11.975 272.052,26.8479\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path d=\"M 272.052,26.8479 297.634,41.7208\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path d=\"M 248.269,26.5416 266.177,36.9526\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1;stroke-dasharray:6,6\"/>\n",
       "<path d=\"M 266.177,36.9526 284.084,47.3636\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1;stroke-dasharray:6,6\"/>\n",
       "<path d=\"M 232.46,11.9327 206.336,26.9631\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path d=\"M 206.336,26.9631 180.211,41.9935\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path d=\"M 230.452,26.573 212.165,37.0943\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1;stroke-dasharray:6,6\"/>\n",
       "<path d=\"M 212.165,37.0943 193.878,47.6156\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1;stroke-dasharray:6,6\"/>\n",
       "<path d=\"M 297.634,41.7208 297.634,108.552\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path d=\"M 285.945,51.7455 285.945,98.5273\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1;stroke-dasharray:6,6\"/>\n",
       "<path d=\"M 300.62,46.7444 326.475,31.3751\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path d=\"M 326.475,31.3751 352.329,16.0058\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path d=\"M 294.647,36.6972 320.502,21.3279\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path d=\"M 320.502,21.3279 346.357,5.95859\" style=\"fill:none;fill-rule:evenodd;stroke:#FF0000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path d=\"M 239.465,142.37 180.211,108.279\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path d=\"M 236.406,127.125 194.928,103.262\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1;stroke-dasharray:6,6\"/>\n",
       "<path d=\"M 239.465,142.37 297.634,108.552\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path d=\"M 242.315,127.193 283.034,103.52\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1;stroke-dasharray:6,6\"/>\n",
       "<path d=\"M 180.211,41.9935 180.211,108.279\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path d=\"M 191.9,51.9364 191.9,98.3364\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1;stroke-dasharray:6,6\"/>\n",
       "<path d=\"M 180.211,41.9935 122.309,7.09091\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path d=\"M 165.492,46.7685 124.96,22.3366\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1;stroke-dasharray:6,6\"/>\n",
       "<path d=\"M 180.211,108.279 122.309,143.182\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path d=\"M 165.492,103.504 124.96,127.936\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1;stroke-dasharray:6,6\"/>\n",
       "<path d=\"M 122.309,7.09091 63.3284,41.1818\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path d=\"M 119.311,22.3241 78.0246,46.1877\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1;stroke-dasharray:6,6\"/>\n",
       "<path d=\"M 122.309,143.182 63.3284,109.091\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path d=\"M 119.311,127.949 78.0246,104.085\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1;stroke-dasharray:6,6\"/>\n",
       "<path d=\"M 63.3284,41.1818 63.3284,109.091\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1\"/>\n",
       "<path d=\"M 75.0167,51.3682 75.0167,98.9045\" style=\"fill:none;fill-rule:evenodd;stroke:#000000;stroke-width:2px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1;stroke-dasharray:6,6\"/>\n",
       "<text style=\"font-size:15px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;text-anchor:start;fill:#FF0000\" x=\"232.46\" y=\"15.4026\"><tspan>O</tspan></text>\n",
       "<text style=\"font-size:15px;font-style:normal;font-weight:normal;fill-opacity:1;stroke:none;font-family:sans-serif;text-anchor:start;fill:#FF0000\" x=\"349.343\" y=\"14.3182\"><tspan>O</tspan></text>\n",
       "</svg>"
      ],
      "text/plain": [
       "<IPython.core.display.SVG object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "m = suppl[0]\n",
    "draw2D(m)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Morgan Fingerprints"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:29.223353Z",
     "start_time": "2019-08-13T23:13:29.218924Z"
    }
   },
   "outputs": [],
   "source": [
    "from rdkit.Chem import AllChem"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 1. <rdkit.DataStructs.cDataStructs.UIntSparseIntVect at 0x7fdfcac2f080>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:29.976387Z",
     "start_time": "2019-08-13T23:13:29.968972Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<rdkit.DataStructs.cDataStructs.UIntSparseIntVect at 0x7ff9f13990d0>"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mfp = AllChem.GetMorganFingerprint(m, radius=1)\n",
    "mfp"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "By using the function <font color='red'>GetNonzeroElements()</font>, we will get a list of Morgan Fingerprints, as shown below"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:31.345479Z",
     "start_time": "2019-08-13T23:13:31.338518Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[10565946,\n",
       " 98513984,\n",
       " 368713488,\n",
       " 864942730,\n",
       " 951226070,\n",
       " 2360741695,\n",
       " 2759434019,\n",
       " 3189457552,\n",
       " 3217380708,\n",
       " 3218693969,\n",
       " 4197146180]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(mfp.GetNonzeroElements().keys())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "After getting a good understanding of what Mol2Vec is and how Morgan Fingerprints work, let's start by importing some modules:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Load trained model\n",
    "Model training are done using ref: [Mol2Vec](https://github.com/samoturk/mol2vec)\n",
    "- Model training is replying on Word2Vec implementation of [gensim](https://radimrehurek.com/gensim/models/word2vec.html)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:33.849136Z",
     "start_time": "2019-08-13T23:13:33.398857Z"
    }
   },
   "outputs": [],
   "source": [
    "from gensim.models import Word2Vec\n",
    "\n",
    "model = Word2Vec.load('db/model_300dim.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### start matching"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In [mol2vec/features.py](https://github.com/samoturk/mol2vec/blob/master/mol2vec/features.py) there are two different functions that calculate Morgan Fingerprints: <font color='red'>mol2sentence()</font> and <font color='red'>mol2alt_sentence()</font>\n",
    "They are the same, except \n",
    "- <font color='red'>mol2alt_sentence()</font> only returns the alternating sentences, which is the list of sentences for all radii combined. \n",
    "\n",
    "Whereas, \n",
    "- <font color='red'>mol2sentence()</font> returns an additional list of sentences for each radius.\n",
    "\n",
    "In this project, we are using <font color='red'>mol2alt_sentence()</font> since the alternating sentence is the list we need."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font color='red'>mol2alt_sentence()</font>: a function that \n",
    "- parameters \n",
    "    - a **mol**\n",
    "    - a **radius**\n",
    "- calculates ECFP (**Morgan Fingerprints**)\n",
    "- returns identifiers of substructions as sentences (strings)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:35.350996Z",
     "start_time": "2019-08-13T23:13:35.247989Z"
    }
   },
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'ms' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-16-754753198383>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;31m# ms is a molecule list\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m \u001b[0msentence\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmol2alt_sentence\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mms\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\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;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m: name 'ms' is not defined"
     ]
    }
   ],
   "source": [
    "# ms is a molecule list\n",
    "sentence = mol2alt_sentence(ms[0], 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### vectorlizing a molecule\n",
    "1. declare a variable called **vecNode**: sum of the vectors of all substructures\n",
    "2. for i in range of length of sentence: however many substructures there are\n",
    "3. a variable vec is the vector of the substructure, get it using model.wv\n",
    "4. cumulatively adds vec to vecNode\n",
    "5. will output a summed vector of all substructures of a mol (output below)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### mol2vec algorithm #1\n",
    "<font color='red'>mol2vec_duplicates()</font> is a function that:\n",
    "- takes in parameters:\n",
    "    - **mol**: a molecule\n",
    "    - **radius**: radius\n",
    "- returns a vector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:40.146848Z",
     "start_time": "2019-08-13T23:13:40.139531Z"
    }
   },
   "outputs": [],
   "source": [
    "def mol2vec_duplicates(mol, radius):\n",
    "    # get sentence\n",
    "    sentence = mol2alt_sentence(mol, radius)\n",
    "\n",
    "    # calculate vecNode\n",
    "    vecNode = 0\n",
    "    for i in range(len(sentence)):\n",
    "        vec = model.wv[sentence[i]]\n",
    "        vecNode += vec\n",
    "\n",
    "    return vecNode"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### mol2vec algorithm #2\n",
    "<font color='red'>mol2vec_unique()</font> is a function that:\n",
    "- takes in parameters:\n",
    "    - **mol**: a molecule\n",
    "    - **radius**: radius\n",
    "- returns a vector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:40.766166Z",
     "start_time": "2019-08-13T23:13:40.758903Z"
    }
   },
   "outputs": [],
   "source": [
    "def mol2vec_unique(mol, radius):\n",
    "    # get sentence\n",
    "    sentence = list(set(mol2alt_sentence(mol, radius)))\n",
    "\n",
    "    # calculate vecNode\n",
    "    vecNode = 0\n",
    "    for i in range(len(sentence)):\n",
    "        vec = model.wv[sentence[i]]\n",
    "        vecNode += vec\n",
    "\n",
    "    return vecNode"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### mol2vec algorithm #3\n",
    "<font color='red'>mol2vec_average_dup()</font> is a function that:\n",
    "- takes in parameters:\n",
    "    - **mol**: a molecule\n",
    "    - **radius**: radius\n",
    "- returns a vector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:41.406180Z",
     "start_time": "2019-08-13T23:13:41.398928Z"
    }
   },
   "outputs": [],
   "source": [
    "def mol2vec_average_dup(mol, radius):\n",
    "    # get sentence\n",
    "    sentence = mol2alt_sentence(mol, radius)\n",
    "\n",
    "    # calculate vecNode\n",
    "    vecNode = 0\n",
    "    for i in range(len(sentence)):\n",
    "        vec = model.wv[sentence[i]]\n",
    "        vecNode += vec\n",
    "\n",
    "    return vecNode/len(sentence)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### mol2vec algorithm #4\n",
    "<font color='red'>mol2vec_average_uni()</font> is a function that:\n",
    "- takes in parameters:\n",
    "    - **mol**: a molecule\n",
    "    - **radius**: radius\n",
    "- returns a vector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:42.012591Z",
     "start_time": "2019-08-13T23:13:42.008171Z"
    }
   },
   "outputs": [],
   "source": [
    "def mol2vec_average_uni(mol, radius):\n",
    "    # get sentence\n",
    "    sentence = list(set(mol2alt_sentence(mol, radius)))\n",
    "\n",
    "    # calculate vecNode\n",
    "    vecNode = 0\n",
    "    for i in range(len(sentence)):\n",
    "        vec = model.wv[sentence[i]]\n",
    "        vecNode += vec\n",
    "\n",
    "    return vecNode/len(sentence)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "***"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Similarity distance #1: Euclidean\n",
    "To calculate the similarity between two vectors, we are using the Euclidean distance. The function <font color='red'>mol2vec_euclidean_similarity()</font> is a function that:\n",
    "- takes in parameters:\n",
    "    - **ms**: a small mol database\n",
    "    - **idx**: index of the query\n",
    "    - **radius**: radius to compute\n",
    "    - **f**: mol2vec function to use\n",
    "- returns a list of Euclidean distances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:43.679609Z",
     "start_time": "2019-08-13T23:13:43.669396Z"
    }
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "def mol2vec_euclidean_similarity(ms, idx, radius, f):\n",
    "    # query, vec\n",
    "    query = f(ms[idx], radius)\n",
    "    \n",
    "    # declare dist list \n",
    "    dist = []\n",
    "\n",
    "    for i in range(len(ms)):\n",
    "        target = f(ms[i], radius)\n",
    "        \n",
    "        # compute distance \n",
    "        distance = np.linalg.norm(query-target)\n",
    "        dist.append((i, distance))\n",
    "\n",
    "    # sort \n",
    "    dist.sort(key=lambda x:x[1])\n",
    "\n",
    "    return dist"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Similarity distance #2: Cosine\n",
    "To calculate the similarity between two vectors, we are using the cosine distance. The function <font color='red'>mol2vec_cosine_similarity()</font> is a function that:\n",
    "- takes in parameters:\n",
    "    - **ms**: a small mol database\n",
    "    - **idx**: index of the query\n",
    "    - **radius**: radius to compute\n",
    "    - **f**: mol2vec function to use\n",
    "- returns a list of cosine distance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:44.976161Z",
     "start_time": "2019-08-13T23:13:44.708666Z"
    }
   },
   "outputs": [],
   "source": [
    "from sklearn.metrics.pairwise import cosine_similarity, cosine_distances\n",
    "\n",
    "def mol2vec_cosine_similarity(ms, idx, radius, f):\n",
    "    # query, vec\n",
    "    query = f(ms[idx], radius).reshape(1, -1)\n",
    "    \n",
    "    # declare dist list \n",
    "    dist = []\n",
    "\n",
    "    for i in range(len(ms)):\n",
    "        target = f(ms[i], radius).reshape(1, -1)\n",
    "        \n",
    "        # compute distance \n",
    "        distance = cosine_distances(query, target)[0][0]\n",
    "        dist.append((i, distance))\n",
    "\n",
    "    # sort \n",
    "    dist.sort(key=lambda x:x[1])\n",
    "\n",
    "    return dist"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To compare the rankings generate by different similarity search, we first conduct a method called **recall**. <font color='red'>compare_rankings_recall()</font> is a function that:\n",
    "- takes in parameters:\n",
    "    - **gt**: ground truth rankings\n",
    "    - **tar**: target rankings\n",
    "    - **k**: k nearest neighbors\n",
    "- returns the proportion"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:45.597153Z",
     "start_time": "2019-08-13T23:13:45.588025Z"
    }
   },
   "outputs": [],
   "source": [
    "def compare_rankings_recall(gt, tar, k):\n",
    "    # get idx from both ground truth and target lists\n",
    "    gt_keys = [gt[i][0] for i in range(1, k+1)]\n",
    "    tar_keys = [tar[j][0] for j in range(1, k+1)]\n",
    "    \n",
    "    # declare a var count to keep track\n",
    "    count = 0\n",
    "    for x in range(k):\n",
    "        if tar_keys[x] in gt_keys:\n",
    "            count+=1\n",
    "    \n",
    "    return count/k"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "***"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Tanimoto vs Euclidean similarity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:47.650731Z",
     "start_time": "2019-08-13T23:13:47.637389Z"
    }
   },
   "outputs": [],
   "source": [
    "def tanimoto_euclidean(moldb, k, radius):\n",
    "    lst1, lst2, lst3, lst4 = [], [], [], []\n",
    "    query_size = 20\n",
    "    \n",
    "    for i in range(query_size):\n",
    "        gt_rankings = tanimoto_similarity(moldb, i)\n",
    "        dup_rankings = mol2vec_euclidean_similarity(moldb, i, radius, mol2vec_duplicates)\n",
    "        uni_rankings = mol2vec_euclidean_similarity(moldb, i, radius, mol2vec_unique)\n",
    "        avg_dup_rankings = mol2vec_euclidean_similarity(moldb, i, radius, mol2vec_average_dup)\n",
    "        avg_uni_rankings = mol2vec_euclidean_similarity(moldb, i, radius, mol2vec_average_uni)\n",
    "\n",
    "        lst1.append(compare_rankings_recall(gt_rankings, dup_rankings, k))\n",
    "        lst2.append(compare_rankings_recall(gt_rankings, uni_rankings, k))\n",
    "        lst3.append(compare_rankings_recall(gt_rankings, avg_dup_rankings, k))\n",
    "        lst4.append(compare_rankings_recall(gt_rankings, avg_uni_rankings, k))\n",
    "    \n",
    "    return [lst1, lst2, lst3, lst4]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Tanimoto vs Cosine similarity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:49.018199Z",
     "start_time": "2019-08-13T23:13:49.004392Z"
    }
   },
   "outputs": [],
   "source": [
    "def tanimoto_cosine(moldb, k, radius):\n",
    "    lst1, lst2, lst3, lst4 = [], [], [], []\n",
    "    query_size = 20\n",
    "\n",
    "    for i in range(query_size):\n",
    "        gt_rankings = tanimoto_similarity(moldb, i)\n",
    "        dup_rankings = mol2vec_cosine_similarity(moldb, i, radius, mol2vec_duplicates)\n",
    "        uni_rankings = mol2vec_cosine_similarity(moldb, i, radius, mol2vec_unique)\n",
    "        avg_dup_rankings = mol2vec_cosine_similarity(moldb, i, radius, mol2vec_average_dup)\n",
    "        avg_uni_rankings = mol2vec_cosine_similarity(moldb, i, radius, mol2vec_average_uni)\n",
    "\n",
    "        lst1.append(compare_rankings_recall(gt_rankings, dup_rankings, k))\n",
    "        lst2.append(compare_rankings_recall(gt_rankings, uni_rankings, k))\n",
    "        lst3.append(compare_rankings_recall(gt_rankings, avg_dup_rankings, k))\n",
    "        lst4.append(compare_rankings_recall(gt_rankings, avg_uni_rankings, k))\n",
    "    \n",
    "    return [lst1, lst2, lst3, lst4]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "***"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Pandas boxplot\n",
    "Import the pandas library"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:50.552962Z",
     "start_time": "2019-08-13T23:13:50.548533Z"
    }
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create the dataframe and plot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:51.263506Z",
     "start_time": "2019-08-13T23:13:51.245741Z"
    }
   },
   "outputs": [],
   "source": [
    "def create_df_bp(moldb, k_lst, radius):\n",
    "    keys = ['euclidean: gt vs dup', 'euclidean: gt vs uni',\n",
    "       'euclidean: gt vs avg_dup', 'euclidean: gt vs avg_uni', \n",
    "       'cosine: gt vs dup', 'consine: gt vs uni', \n",
    "       'cosine: gt vs avg_dup', 'cosine: gt vs avg_uni']\n",
    "    df_lst = []\n",
    "    for k in range(len(k_lst)):\n",
    "        eucli, cos = [], []\n",
    "        for i in range(4):\n",
    "            eucli.append(tanimoto_euclidean(moldb, k_lst[k], radius)[i])\n",
    "        for i in range(4):\n",
    "            cos.append(tanimoto_cosine(moldb, k_lst[k], radius)[i])\n",
    "        values = eucli + cos\n",
    "        d = dict(zip(keys, values))\n",
    "        df = pd.DataFrame(d)\n",
    "        df_lst.append(df)\n",
    "    \n",
    "    # plot, subplots\n",
    "    fig, axes = plt.subplots(nrows=len(k_lst), ncols=1, sharex=True)\n",
    "    \n",
    "    for k in range(len(k_lst)):\n",
    "        ax = df_lst[k].boxplot(ax=axes[k], rot=-90)\n",
    "        ax.set_title(\"recall @ \" + str(k_lst[k]) + \" for radius=\" + str(radius))\n",
    "        ax.set_ylabel(\"recall @ \" + str(k_lst[k]))\n",
    "        ax.set_ylim([0, 1.1]) \n",
    "    \n",
    "    return plt.subplots_adjust(hspace=0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:51.760441Z",
     "start_time": "2019-08-13T23:13:51.757938Z"
    }
   },
   "outputs": [],
   "source": [
    "#create_df_bp(c_small_moldb100, [1, 5, 10], radius=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#create_df_bp(c_small_moldb200, [1, 5, 10], radius=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#create_df_bp(c_small_moldb500, [1, 5, 10], radius=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-07-25T23:26:48.612300Z",
     "start_time": "2019-07-25T23:26:48.610036Z"
    }
   },
   "outputs": [],
   "source": [
    "#create_df_bp(c_small_moldb1000, [1, 5, 10], radius=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "***"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:58.008528Z",
     "start_time": "2019-08-13T23:13:57.999312Z"
    }
   },
   "outputs": [],
   "source": [
    "def combine_8funcs(k):\n",
    "    # put 8 functions together for all sample sizes\n",
    "    k_100 = tanimoto_euclidean(c_small_moldb100, k, 1)\\\n",
    "    +tanimoto_cosine(c_small_moldb100, k, 1)\n",
    "    k_200 = tanimoto_euclidean(c_small_moldb200, k, 1)\\\n",
    "    +tanimoto_cosine(c_small_moldb200, k, 1)\n",
    "    k_500 = tanimoto_euclidean(c_small_moldb500, k, 1)\\\n",
    "        +tanimoto_cosine(c_small_moldb500, k, 1)\n",
    "    k_1000 = tanimoto_euclidean(c_small_moldb1000, k, 1)\\\n",
    "        +tanimoto_cosine(c_small_moldb1000, k, 1)\n",
    "    \n",
    "    return k_100, k_200, k_500, k_1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:58.281662Z",
     "start_time": "2019-08-13T23:13:58.278111Z"
    }
   },
   "outputs": [],
   "source": [
    "# takes in a list of lists, returns a list of meds of all lists\n",
    "def med(lst):\n",
    "    med_lst = []\n",
    "    \n",
    "    for i in lst:\n",
    "        med = np.median(i)\n",
    "        med_lst.append(med)\n",
    "    \n",
    "    return med_lst"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:13:58.914406Z",
     "start_time": "2019-08-13T23:13:58.908197Z"
    }
   },
   "outputs": [],
   "source": [
    "def create_df_line(lst, k):\n",
    "    cols = ['euclidean: gt vs dup', 'euclidean: gt vs uni',\n",
    "       'euclidean: gt vs avg_dup', 'euclidean: gt vs avg_uni', \n",
    "       'cosine: gt vs dup', 'consine: gt vs uni', \n",
    "       'cosine: gt vs avg_dup', 'cosine: gt vs avg_uni']\n",
    "    df = pd.DataFrame(lst, index=['100', '200', '500', '1000'], columns=cols)\n",
    "    line = df.plot.line(title=\"recall vs sample size: k=\"+str(k))\n",
    "    \n",
    "    return line"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:14:01.858334Z",
     "start_time": "2019-08-13T23:14:01.637295Z"
    }
   },
   "outputs": [],
   "source": [
    "ls = [':', '-', '--', '-.']\n",
    "fig, ax = plt.subplots()\n",
    "df.plot(y=0, kind='line', ax=ax, linestyle=ls[0], linewidth=1)\n",
    "df.plot(y=1, kind='line', ax=ax, linestyle=ls[0], linewidth=5)\n",
    "df.plot(y=2, kind='line', ax=ax, linestyle=ls[2], linewidth=3, alpha=0.5)\n",
    "df.plot(y=3, kind='line', ax=ax, linestyle=ls[3], linewidth=5, alpha=0.5)\n",
    "df.plot(y=4, kind='line', ax=ax, linestyle=ls[0], linewidth=7, alpha=0.5)\n",
    "df.plot(y=5, kind='line', ax=ax, linestyle=ls[2], alpha=0.3)\n",
    "df.plot(y=6, kind='line', ax=ax, linestyle=ls[2])\n",
    "df.plot(y=7, kind='line', ax=ax, linestyle=ls[3], linewidth=5, alpha=0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:17:47.167924Z",
     "start_time": "2019-08-13T23:14:19.779931Z"
    }
   },
   "outputs": [],
   "source": [
    "lst = combine_8funcs(5)\n",
    "# find median\n",
    "k_5_100_m = med(lst[0])\n",
    "k_5_200_m = med(lst[1])\n",
    "k_5_500_m = med(lst[2])\n",
    "k_5_1000_m = med(lst[3])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:18:47.899764Z",
     "start_time": "2019-08-13T23:18:47.509739Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7ff9f1d5e748>"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "values_5 = [k_5_100_m, k_5_200_m, k_5_500_m, k_5_1000_m]\n",
    "create_df_line(values_5, 5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:24:43.324439Z",
     "start_time": "2019-08-13T23:18:50.562856Z"
    }
   },
   "outputs": [],
   "source": [
    "lst_10 = combine_8funcs(10)\n",
    "# find median\n",
    "k_10_100_m = med(lst_10[0])\n",
    "k_10_200_m = med(lst_10[1])\n",
    "k_10_500_m = med(lst_10[2])\n",
    "k_10_1000_m = med(lst_10[3])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:25:02.625294Z",
     "start_time": "2019-08-13T23:25:02.143077Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7ff9f1d3a550>"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYcAAAEICAYAAAC0+DhzAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjAsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+17YcXAAAgAElEQVR4nOydd3hURReH39mS3ggdAiT03pvyKYg0kSrSURApgogNRUCRKk1REAFBEJVeBEKRjoCg0hQUCCUhQEhCC+nZzZb5/tgFQwpJYEMCmfd59mHvvVPOvWzub86UM0JKiUKhUCgUKdHktgEKhUKhyHsocVAoFApFGpQ4KBQKhSINShwUCoVCkQYlDgqFQqFIgxIHhUKhUKRBiYPCoQghxgkhltq/+wshpBBCl9t25SRCiGZCiDAHl/mMEOKsI8tMp45+QojfcrIOxeOLEgeFIg8ipTwgpayU23ZkhhCiuBAiUAgRbm8I+Ke67iyEWCyEiBVCRAoh3ssdSxXZRYlDPuRJb8krHilWYBvQJYPr44AKQBngOeBDIUSbR2Oa4mFQ4pBPEEKECiFGCiFOAglCCJ0QooQQYp0Q4oYQ4qIQYniK9FohxGghRLAQIk4IcUwIUcp+bZYQ4oq9NXhMCPHMA9jzkRBibapzs4QQs+3f+wkhQux1XxRC9M6gnIZCiKN2W64JIWamuLbG3lqNEULsF0JUS3FtiRBirhDiFyFEvBDioBCimBDiKyHEbSFEkBCiTqrnN0oIcdp+/XshhEsGNmX4XNNJ29ZeZpwQ4qoQYoT9/N2uKiFEd7uNdz5GIcSv9mvOQojPhRCX7fc/XwjhmoX/gvRsmSGE+E0I4Z3VPFLKa1LKucCRDJK8CkyUUt6WUp4BFgL9HsQ+xaNFiUP+oifwIuCDrcW3CTgBlASeB94RQrS2p33Pnr4t4AX0BxLt144AtQFfYDmwJqMX5X1YAbQVQniBTYyAbsByIYQ7MBt4QUrpCTwN/J1BObOAWVJKL6AcsDrFtV+wtVqLAMeBZanydgM+BgoBRuB3e7pCwFpgZqr0vYHW9noq2vPegxBCw/2fa2oWAYPt91kd2JM6gZRylZTSQ0rpAZQAQrA9P4BpdltqA+XtdY5NYU+0EOJ/GdR912YhxEKgJtBKShkjhPifPW9Gn/uWaS+3gN3eEylOnwCqpZ9DkaeQUqpPPvgAoUD/FMeNgMup0owCvrd/Pwt0zGLZt4Fa9u/jgKX27/6ABHQZ5PsNeNX+vSUQbP/uDkRj66pwzaTu/cB4oFAm6Xzstnjbj5cAC1Ncfws4k+K4BhCd6vm9keK4bQp7mwFhWXmu6dh1GRgMeKU6f7fMFOc0wGZgnv1YAAlAuRRpngIuZvH/rR/wJ7AKWAc4PcTvS2d/vv4pzpWyn3NJca4lEJrbfw/qk/lHeQ75iyspvpcBSqRsDQKjgaL266WA4PQKEUK8L4Q4Y++uiQa8sbW2s8tybN4JQC/7MVLKBKA78AYQIYTYIoSonEEZr2NrOQcJIY4IIdrZbdQKIabau8Visb3cSWXntRTfk9I59khVV8rndwlbqzg1mT3X1HTBJjSXhBD7hBBPZZAOYDLgCdzppioMuAHHUtS1zX4+q5QHOgLjpZTJ2ciXFeLt/3qlOOcFxDm4HkUOoMQhf5EyBO8VbC1MnxQfTyll2xTXy6UuwD6+MBJbl0wBKaUPEIOtFZtd1gDNhBB+QGfs4gAgpdwupWwJFAeCsPVVp70hKc9LKXti6zqaBqy1d0v1wvbSa4FNvPzv3MID2HmHUim+lwbC00mT2XNNbf8RKWVHu/0buLdb7C5CiB7YhPRlKaXJfvomNhGrlqIub2nrfsoqZ4DXgF+EEHdnRwnbVNr4+3wyHWeSUt4GIoBaKU7XAk5lwz5FLqHEIf9yGIgVtkFqV3tLu7oQooH9+nfARCFEBWGjphCiILaWqxm4AeiEEGO5t2WYZaSUN4Bfge+xvVDPAAghigohOthf8kZsLVBLemUIIfoIIQpLKa3YuqKwp/W0572FrXX92YPYmIo3hRB+QghfbN7AqnTSZPZcU9ruJIToLYTwtr/wY0nnPu0D418DnezPDAD7PS8EvhRCFLGnLXmf8Y10kVKusN/PLiFEOfu5A9I+zpHB50AK+1wAZ/uhc6rxpx+Bj4UQBeze30BsXXqKPI4Sh3yKlNICtMc2kHkRWyv0O2ytbLANxq4GdmB7aS0CXIHt2AZ6z2HrWjFwb3dLdlmOrXW/PMU5DfA+tpZ5FNAUGJpB/jbAKSFEPLbB6R5SSgO2l9Il4CpwGvjjIWxMaesObAPCIcCk1Amy8FxT8woQau/6egPok06ajkAB4LcULfdf7NdGAheAP+xl7AJSegBZbeX/AEwA9ohUaxWyQBL/dSEF2Y/v8Cm27slLwD5ghpRyWzbLV+QCQkq12Y9CkRlCiFBggJRyV27bolA8CpTnoFAoFIo0KHFQKBQKRRpUt5JCoVAo0qA8B4VCoVCkIc8FYCtUqJD09/fPbTMUCoXiseLYsWM3pZTZWQB5X/KcOPj7+3P06NHcNkOhUCgeK4QQlxxZnupWUigUCkUalDgoFAqFIg1KHBQKhUKRBiUOCoVCoUiDEgeFQqFQpCFL4iCEaCOEOCuEuCCE+CiDNN2EbbvDU0KI5SnO9xVCnLd/+jrKcIVCoVDkHJlOZbVv3/gNth2cwoAjQohAKeXpFGkqYNvtqomU8naK8MG+2KIy1se2l8Axe97bjr8VhUKhUDiKrKxzaAhckFKGAAghVmILIXw6RZqBwDd3XvpSyuv2862BnVLKKHvendhCLK9AoVDkGW6EXmbjN8uRetB56dC46x9uWyTFY09WxKEk98brD8O2T25KKgIIIQ4CWmCcPWZ7enlLpq5ACDEIGARQunTprNquUCgcwL5FS/gz9CqJ7iabf5/EvTsyoOKv5UeyIg7ptR9S/1p0QAVsm6L7AQeEENWzmBcp5QJgAUD9+vXVL1GheATEXb/GT59/yXU3FzyFC36JLpx3OU+9mMqU05THU+eBm1YLgFVK4s0mkrVGnIrqKdwkAN8GAWjt1xW5z/gJExxaXlbEIYx79871I+3euWHAH/atDi8KIc5iE4swbIKRMu+vD2qsQqFwDAeWfM++cxcwu7lQxVSSC65xvDvuHU5eu8xrWwdh0l6nW5kRDC/WjGt7zmO4kogWJ3yEB7obAuOGCC6uu0qi1QBuFjwr+1C8dXVcCj3QjrGKPEimIbuFEDpsW0I+j23LxSNALynlqRRp2gA9pZR9hRCFgL+wbZMogWNAXXvS40C9O2MQ6VG/fn2pYispFDlDbEQEP335OTfcPPG0utAouSIrip7gi2FjcNE7ARAWE0WXnweSqDlHY59X+bb9+2g0tomN5sRkInf9S/SJa1hiwBUX3HS2NuYd78KkMeJUVEfBp/0p2Kic8i4eEUKIY1LK+g4rLyv7OQgh2gJfYRtPWCylnCyEmAAclVIGCiEE8AW2wWYLMFlKudKetz+2zcuxn//+fnUpcVAocoYDSxaz99x5rHonqlpKUSm5DJP91vPtgOkUcPO4J22cMYlOq4dz3foHZZ1asabrNJx06Xc0xJwN59rusyRdTkRrcsJD54xO2HqUjVYrCRabd+FRwYcSraviUtQnx+81P5Ir4vAoUeKgUDiW2IgIfvzyc266eeJsghbWugirCyPLfMeiHnPw9y2Sbj6zxULPdZ8QlLQJX+qysdtcfFzdM63PnGQics8pov+KxHJb4iJccE/hXSSYTSRrjOiL6CjUqDQFm1RQ3oUDUOKgUCiyzL7F37HvQjBWvROljV40px6hxDGy7Czmt1tInRIBmZbx1pbZ7L3xHa7WANZ0WpihmNyP2PORXNsdROKlBLTJGXgXrmY8yvtQok1VXIoVyHYd+R0lDgqFIlOir4bz01dfcMvdE63RwNOa+lSyFOWoJoJxZWcyo+l8WlaoneXypu9fzY/BU9BZfVnY+lsa+JV/KPvMRhPX9pzm9vEIm3fBf96FlJJ4i5lkYUBfWEfBhn4UeqaS8i4yQYmDQqG4L/sWLWRfcAhWvRPFkgw87dQWX7OGbU4X+KrMbEbV/5LetZplu9xlJ35l6rEPAR2TGn9Jx6qplzs9HHEh14jcFUTixXg0yXo8tC7oNTbvItlqJcFswOpmxqOcF8VbV8ethPIuUqLEQaFQpMvtq2Es/Womt9y90BqTaFq2FkVDSuNqtbLK8x9+8PuW/uXH8d7/XnrgOvaG/MPbe4dh1cQzpMo43mzc3nE3kApzspnre08RdSwCS5TEGRc8UnsXGNAV0lKwQUkKP1MZrVOe29zykaHEQaFQpGHvwm85cDEUq96J4oYEXmj/BsmbI9BKyfeFj7O+8CJeKPYW01sPeui6zlwPo/emgSRrw+lQYjiftXrdAXeQNeIv3SByZxAJIXFoDDo8dKm8C4sRq4sJ97JeFG9TDfeSvo/MttxGiYNCobjL7cuX+enrWUS5e6I1JNG6Tm1KFGmCaVsoFiFYVOY4m92+o55XD5Z0HuOweiPjbvPS2jeI05ymjmd3lnQafXctxKPEkmzm+r4zRB0Lx3zTgjMuuGt1CCGQUpJgMWO0exe+9UtS5Nkn17tQ4qBQKADY/e08Dl6+jFXnRAljIn1GjiZizw30hyNJ0mhYWv0MG8yzCXBqyYbunzv85Z1oMtJp1TtEWH6jlK4ZP3f74u5Cutwk/spNInecISEkFo1Bj7vWGSf7vZusknizAaurCXd/T4q1qoJHmcK5bLFjUOKgUORzbl0KZdmc2US5e6EzJNG6Xl3qdenKue9O4RESTaxew+bGV/jp5mQKUp8dvRdmuIDtYbFarbyyfgIn49fhLWuw/uX5FPbIWyE0LMlmbhw4y60jYWm8C4B4swkjRrS+goL1SlDkuaqPpXehxEGhyMfsnj+Xg1fCsOr0lDQm0XvkKJw9vQn66jjeN5OI8dBz6LnbzDo/CndZjh09f8LbxS3H7RqxbT7bIufibC3N8vYLqFS4RI7X+TAkXI0icscp4oNj0STpcNe63OtdWAxYnUy4BbhTvGVVPAKyv7bjUaPEQaHIh9y6FMrSObO5bfcW2jSoR/0u3TAlJHPui+N4J5qIKerG+ZbJfHzkLfTWgmx6eQV+3o9uQParQxv47uwEtFYv5jSfyzMBVR9Z3Q+LxWLh5m9nufVnGKYbZpykCx4pvIsEsxkDBrQFBAXqFqdo86ronPW5bPW9KHFQKPIZu+bN4VBYuM1bSE6i90djcPP2JvF6Ipdm/4WHyUJi5YJENdcxaHc/hHRiRbulVCtaKvPCHcy6U4cY9+d7gOSTBl/Qrcb/HrkNjsIQeZvwbaeJvxCNSNLhpnXBOYV3kWAxYnFKxrWMO8Wer4xXhWK5aq8SB4Uin3Dr4kV++uZroj280BkSeaFRI+p17gJA9IVori/6B2erxNqkJManXOge2BurSOKb5xbzbEC1XLP70KUghuwagkUTw+sVP+HdJp1zzRZHYrFYuHXwPDf/vIzpuhknqzPuOj2alN6FNKAtAD51ilOsRbVH6l0ocVAo8gE75nzNHxERWHV6/JKT6GX3FgCuHbtG/OpzCCTO7cuhreVG2zW9SRaRjKv/NS/XaJLL1sP5mxH0DByIQXOZVkWHMPOFIbltUo5guBZN+PbTxJ+PRiRq7/EuzFISbzZi0SfjWsaNoi0q412heI7ZosRBoXiCuRl8gZ/mzyXG3eYttG3cmLqd/lvRfHnnJay7LmESggKvVMG5vDutlvclTpxlSOVJObpiObvcSoyj0+ohRIsTVHfrzLIu43JlLcSjxGKxcOvPYG4dCiX5mhm91RmPFN5FotlMkjSg9QGfWkUp1qI6OjfHTP9V4qBQPKFsnzObPyIikTodpUxG+owag7Pnf9NCz688i/Nf10jUaij5Zi1ci7nRatlgbsg/6VTyfSa26Jd7xmeAwZRMlzUfcNm0h+KaJqzvNgt3Z+fcNuuRYrgZS8T2f4kLioZELe4ZeRel3CjavALeVfweqB4lDgrFE8b18+dY9u18Yjy80Ccl0vbpp6jT8b9+eqvVytlv/8HzUiyxTlrKvlcXJy8nXlo9kmDjNp4u0JdvO4zIxTu4P1arldc3TuVo7Ao8rFVY//K3FPPMv0HzLBYLUUcucuPgRZKvmXCypPIuLBaSrElovMGnZhGKtayK3t0103Jzaye4NsAsbDvBfSelnJrqej9gBrZtRAHmSCm/s1+zAP/Yz1+WUna4X11KHBT5iW1ff8WfkdeROh2lTUZ6p/IWLMkWgr48jvdtAzFezlQaURedk44BG6fxZ/RSKrq8yJqunz0W3TVjdi5m49VZOFlL8NOLC3JlNlVexXgrjogdp4g7E4VM0OCmccVFa/s/tdi9C7M+GRc/V4o2r4BP1bTP7pGLgxBCi20P6ZZAGLY9pHtKKU+nSNMPqC+lHJZO/ngppUfq8xmhxEGRH7h2/hzLU3gL7Z75H7Xa3dtuSo5L5vwXx/A2mIkt6UHlN2uh0WgYs3MxgeFfUlTzFNt6zUP3GO1zMO/wVr45NRaN1Y2vms2hebmauW1SnsRisRB9PJQbv4VgjEhGb3HGQ+d017tIslhIshoQXhLv6oUo3qY6Th5uDhWHrKwRbwhckFKGAAghVgIdgdP3zaVQKNJl66yZHLlxC+nuQRmTgV6fjMXZ4972U3xEAmFz/sLLbCWxRmGq9qkCwNe/b2Tj1Vl4yaps7PG144Th6nHYPQGSbjumvAwYAjS26BnufJvR+3ozY7sbz2hdcrTOxxEtUBAo6AaUs50zGp2JCK9CbGIlECVxc/LGJUkLRwyEHz7icBuyIg4lgSspjsOA9Hb56CKEeBabl/GulPJOHhchxFHADEyVUm5InVEIMQgYBFC6dOlsmK9QPD5Eng1i+XcLiHX3Qm9Kpt1Tz1LrxbSzi6KCori15BSuUmJpWoqKbW1bea46+RvfBo3HWfqxvtsCxwzsmo3w61Q4OAvcC0PxWg9fZibUAZaYDAw2X+Ytt0RGJ3vQzb1gjtf7uOPsAf4FI4CIu+duXSvAjYhyGJMd30WXlW6lrkBrKeUA+/ErQEMp5Vsp0hQE4qWURiHEG0A3KWVz+7USUspwIURZYA/wvJQyOKP6VLeS4klky1czOXrzFlKrxd9soueoMWm8BYCIPyJIWn8BANfO5Sne2DYvfm/IPwz/dQAa6c7ajsuoUMgB8+XDjsHGoXAjCOr0gVaTwdXn4cvNIqFR1+m6YRBJmhCaFurPN+3eeWR1P4k4eswhK6NYYUBKWfIDwlMmkFLeklIa7YcLgXoproXb/w0BfsXWcFAo8gWRZ04zc8R7HImORWdKpkujRvSbPCVdYQjdEkLy+vOYBfi8Vu2uMPwbeZm39w4FtCxo+e3DC4PJADs/hUUtwBgHvddBx28eqTAA+PsWYWfPZRQUddh/axFdVn2E2WJ5pDYoMiYr3UpHgApCiABss5F6AL1SJhBCFJdS3vF1OgBn7OcLAIl2j6IQ0ASY7ijjFYq8isVi4ZdZMzkWFY108yDAnEyvcRPQu6Y/JfHcT2dw/fcGCToNfsPq4FHcHYAr0bd4ZctArJokpjSeR6PSFR7OsLCjsGEI3DwHdV+FVpPAxfvhynwIfFzd2dl7Ed3WjuKcYQutlt1gfbevH0kkWcX9yVQcpJRmIcQwYDu2cZLFUspTQogJwFEpZSAwXAjRAdu4QhTQz569CvCtEMKKzUuZmnKWk0LxJBJ+6l9WfL+YOA8vnEzJdHjuOaq3aZtuWqvVStCcE3iFxxPjqqfC+/Vw8rStmI1OSqDLzwMwaa7xXo3ptK/S4MGNMhlg72T4fQ54loA+66B8iwcvz4E46XT83G0aQ7cU42DUElqt6MPazgsp5aPGIXITtQhOoXAQFouFrV99wfHbMUiNlgBpptfoTzL0FswGM2dnHsc71khMARcqv1sXrZNt9pHBlEyr5QOJkn/RK2AUo5v2fHDDrhyGjW/avIV6/aDlRHDJWxvy3GHC3qWsvvQ5ekthFr+wgDolAnLbpMcGR485PH7bHSkUeZDwf/9h+Q/fE+9u8xY6Pf88VVu1yTC9IdpAyMzjeCdbiCvjRZXBNe4uZLNarXRZ8wG3OU7zwoMeXBhMSbBnEvz+DXj7wSvroVzzByvrETH2uT6UOlaUmSdG03fbK0xtMou2leplnlHhcJQ4KBQPgcViYfPMz/k7Jg7p6kE5i4ke9xlbAIi7EsfVeSfwsFhJqlOEKj0q33O97/rJXDbtoYZ7Z2a/+FYGpWTC5T9tM5FuXYD6/aHF+DzrLaTmtXotKeFViA8OvMWHhwYTETeZ1+u3zm2z8h1KHBSKB+TqyROs+OkHu7dgpFOLFlRtef+X2M1TN7n90xlcpIQWZajQssw919/fNo+/41fjp2vK0pfGZd+o5ESbt/DHXPAuBa9uhLLNsl9OLtO6Qh2Key6l75bBfPnvSMJir/Np81dy26x8hRIHhSKbWCwWNn0xgxOx8TZvwWqix4SJ6J3vv9L36m9XMW4KRoPAtVslitYres/1GQfWsD1yHgWoxfpuM7MfL+nS77axhahgaDAAWowDZ8/slZGHqFnMn00vr6DLzwNZe2U6VzaGs6D9B49FHKknASUOCkU2uHryBCt+/IF4Dy+ckw10bt2Gys9nPuvnYmAwmoNXMWsERQbUwKfcvWsKfji+mx8ufIarDCCwx3xc9NmI8Z+cCHsmwh/zwKcU9N0EAc9m99byJH7evuzqtZROq4fzZ/RSOq26xtqu03HSqVdXTqNmKykUWcDmLUznRFwCUmgoj5XuYz7O1FsAOPv9KdyCbhGv11JmeB3citw7h3/7ub8Y8dsgtNKHDZ2X4e9bJOuGXToEG4bC7YvQYKDdW8hynMvHBrPFQq91YzmTFEgB6hLYbS4+ru65bVaeQs1WUigeMVf+/ouVy34iwd0LZ6OBzm3aUrl55rN+rGYrZ77+C+9ricS4O1Hx/bro3e/1CI6GXWDEb8MQOLOk7cKsC0Nygi1Q3p/fQoEy0HczBDzzILf3WKDTalndbTLDtxRjz42FtFzRi1WdFlDWt2jmmRUPhPIcFIoMsFgsBM6YxsmERKQQVNBAt1FjsuQtmJPMnP3iKN7xJmIKu1L57bpodff2lYdEXeOl9b2wiDi+fHYhLcpnMehd6G+2sYXbodBwMLT4FJzyTyt6+v7V/Bg8BZ3Vl29bzn/4VeNPCMpzUCgeAVf+Os6K5UtJtHsLXdq+SMVmz2Upb+LNJEJnHcfbZCW+nA9VXq+WZhD1VmIc3TYMwKyJYlTdL7MmDMZ42D0eDi+AAgHQbwv4/+9Bbu+x5sNnu1HKuyifHfuAAbteZWLjL+lUtXFum/XEocRBoUiBxWJhw/Qp/JtoQLq6UxEr3SdPQavXZyl/TGgMkd/+g4fVirFhMSp3qZgmTaLJSIfVgzFoLtG//Dh612qWecEX98PGYRB9GRoNgec/yVfeQmp61mpKcc/FDN/7Jh//+SZXY8fxZuO04c8VD44SB4XCzqWjR1i1aiWJ7p64GA10adeOCs82y3L+63/fIHZFEE5INC8EUK5Z2hj7VquVTqveIVb8w4vFhvPe/166f6HGeNj1KRz5DnzLwmtboczT2byzJ5NmZauzymMZvTcPYl7Qx1yJjWRqq4G5bdYTgxIHRb7HYrGwftpn/JtkBFc3KglJt2x4CwBhey9j2hYKQuDRswqFaxVON12PdZ8QYfmNel49mNY6kxdZyD4IHAbRV6Dxm9D8Y3BS0UpTUqWIH790W0HnNYPZEjGbq+si+aHzGLUWwgEocVDkay4d+ZOVq1eTZPcWXu7YkfJNsjfrJ3jdOfSHI0nWaCg2uAbe/umHwH5z81ecSQwkwKklizuOyrhAYxzsHAtHF4NvOei/DUqrPvWMKOrhza4+P9B51bv8Hb+aF1dcZ123z3HTO2CnvHyMmq2kyJdYLBZ+njqZUwYTAJX1WrqOGpMtb8FqtXLuu1N4hEQTq9cQ8G5dXH3Tj6k0Ye9S1lyeRkHqs6P3wowXcYX8Chvfgpgr8NSb8NwY5S1kEavVyqvrJ3Iifi1esgYbXp5PYY/HI56UI1CzlRSKh+TiH3+wet0am7dgSKRr55co93STbJVhMVsJ+vI43reSiPHQU+n9+uhc0/9zWnDkF1Zf+hx3WZGNPb9JXxgMsTZv4dj3ULA89N8OpdPbql2RERqNhqVdPuWD7cX4JeIb2qzuyfL2C6lUuERum/ZYosRBkW+wWCysmzKJ00YzuLhRVQtdPpuaLW8BwJSQzLkvjuOdaCKmmDtVhtVGo0u/j3vj6T+Z/e8n6K3FWN91Yfo7nAXvgcDhEHsVnh4Oz40GfcZRXRX3Z0brwZQ6VIwFZ8fTLbAXc5rP5ZmAqrlt1mNHlkZthBBthBBnhRAXhBAfpXO9nxDihhDib/tnQIprfYUQ5+2fvo40XqHIKiG/H+Lzjz7ktFniakjk1bYv0O2TcdkWhsTriVyYcgSvhGQSKvlS7Z26GQrDoUtBfPzHO2is7ixtt5ASXr73JjDE2EThp842Mei/A1pNVMLgAIY/3ZHxDedgFUaG7n2NVScP5LZJjx2ZjjkIIbTAOaAlEIZtT+meKbf7FEL0A+pLKYelyusLHAXqAxI4BtSTUt7OqD415qBwJBaTiXXTPrN5C0BVZx1dRn2MVqvNdlnRF6K5vugfnK0Sa5OSBHQol2HaoBthdA/sg1UYmPvc4rQt1wu7bMIQF2HzFpqNAn3mK68V2ePQpSCG7BqKRXOb/hU+yXzq8GNMbow5NAQuSClD7AasBDoCWdkLujWwU0oZZc+7E2gDrHgwcxWKrBN86CBr1v+Mwd0TV0Mi3bp0JaDxg836iTwaScKa8+iQ6NqXo+T/SmacNu42vQIHYdHEMr7hnHuFwRAD28fAXz9BoUrw+i7wUzud5RRPl6nMz52W033jQBZfGEdY3DVmvjAkt816LMiKOJQErqQ4DgPSGynrIoR4FpuX8a6U8koGedP8VQkhBgGDAEqXLp01yxWKDLCYTKyZMpkgkwVc3Kim1/DS1OkP5C0AXNpxCbn7ElYhKPBKVQpVK5Rh2uBSWZMAACAASURBVDhjEp3WDiJZe5WhVSbTpVqKBWvnd8Kmt23ewv/ehaYfKW/hEVCuYDG291hOp9VD2Hl9Lj3WhLO8y3i1FiITsvJ0RDrnUvdFbQL8pZQ1gV3AD9nIi5RygZSyvpSyfuHC6S8eUiiywoWDB5gx+iOCrOBqSOS1Du3oOmbsAwvD+ZVBiN2XMGg1FHur9n2FwWyx0HHVMBI0QXQp9R5DG7WzXUiKhg1vwrKXwdkLBuyyhdZWwvDIKOjmyc7eiymjf55TiRtovWwoCUZjbpuVp8mKOIQBKeMA+AHhKRNIKW9JKe886YVAvazmVSgcgcVkYsX4T1m6fQcGF1eqO2kZMXU6ZRo82HRQq9XKmbkncP37BvEuOvxHNsCzZMa7qlmtVl5e8xE35GGeLtCX8c/b516c2w5zG8OJFfDM+zB4H5RU3Ui5gYveicAeM2ng3YtI60FaLH+FyLgMhz/zPVkRhyNABSFEgBDCCegBBKZMIIQonuKwA3DG/n070EoIUUAIUQBoZT+nUDiMCwf2M2PMKM5KgVtSIq917MjLoz95YG/BkmzhzPSjeF6OJcbLmYpjGuLiff/VtoMCZxBs3EZFlxeZ1+49SLoN64fA8m7gWgAG7obnx4JOrdrNTTQaDYs7jaJjyfeIE2d5YXVPTl27knnGfEimYw5SSrMQYhi2l7oWWCylPCWEmAAclVIGAsOFEB0AMxAF9LPnjRJCTMQmMAAT7gxOKxQPi8VkYtXkiZyzSoSzCzWcdXT65MHHFgAMMUaCvzyOt8FMrJ8nVYbWzLRvevTORfwZs5RimqdZ9fJkNOd32MYWEm7Asx/YPkoU8hSTWrxGqcNF+frUWHpu7sXMpnOyvp9GPkGFz1A8lpzf/ytrN2/G6OaBW0Is3Xv2okzdh5vFFx+eQNg3f+FutpJUozAV+1TJNM/sQxtZcG4sXrIKO1/6Cvfdn8LJlVCkGnSaCyVqP5RNipxl05kjjP79bcDMiNpT6Fv3+dw26YFR4TMU+RqT0cDqKZM5bwXh7EItFyc6fDLjobwFgKigKG4tOYWrlFialqJi24BM86w6eYAFZ8fjLP0IrPcy7gufhcRb0HQkPDMCdE6ZlqHIXdpXaUBxzx8ZsP0NZpx4n6txIxndtGdum5UnUJ6D4rEhaM8e1m/bitHNA/eEWHr06kOpOnUfutyI38NJ2hAMgOtLFSjeqFimefaG/MPwXwegk65scS9FiaBAKFoDOn0DxVX3xOPG5egbvPzzQJK0wTzj25+57d/NbZOyjaM9ByUOijyPyWhg1eRJXECDkFZqebrT/v0PH9pbAAjdEoLYH4ZRCAq9Vg3fSr6Z5jkZGUqfLa+gFwZ+vh1NmYQo27jC/95T3sJjTHRSAp1Wv8ktjlHBuS2ru36GzgG/sUeF6lZS5CuCdu9i/fZt/3kLvV+hVO06Din73I+ncT11kwSdBr9hdfAonvm2m1eib/Haltdx0sTwU0Q4ZXwrQ++foVgNh9ikyD18XN3Z0fs7uq0dxXnjVlouu8GGbnPSD5aYD1CegyJPYjIaWDl5IsFoEVYrtb09affeCId4C1arlaA5J/AKjyfGRUeF9+vh5Jl5iz86KYEuyzpwW3eNb65H8VSjd2wrnbXZC96nyNtYrVbe3PIVv0V9j5u1Ams7L6SUT8HcNitTlOegeOI5s2sn63dsJ9nNA4/4WHq+2peSNR3Tj282mDk78xjescnE+LpQ+Z26aJ0yFxxDTCTDV3XkhlMC4xJceOrVHVCsukNsUuQtNBoN89q/x6Rfi7Hy4nTar+vB9y8soE6JzCcpPEkocVDkGUxGAysmTSREaBFOLtRxd6XdJw++mC01hmgDITOP451sIa6MF1UG18hSfB3rv+uZtO8j/vJwoo+oyUtv/KC8hXzAx8164edZhC9OjOLVbX2Y1mQ2bSvln9XtShwUeYLTO7ezYdcukl3d8YiPpVe/1yhR3XH9+HFX4rg67wQeFitJdYpSpUelzDMl3IStI5h3dQ8bC3jT0KkZI3t+7TCbFHmffvVaUNKrEO8feIsPDw3mauxkBjZondtmPRKUOChyFVNSEss/m8hFoUPonajn6UZbB3oLADf/ucntZWdwkRJalKFCyzKZZzq1Hra8z0qdmfkFvfHTPsPC7rMcZpPi8aFlhdos9fyJvlsGM+vUh1yNu8a45q/mtlk5jhqQVuQa/27bSuDevSS7uuOZEEvPfv0pUc2x/fhXD1wleXMwVgQe3SpStF7R+2eIvwFb34fTG9lQpDKfuCXiQy129l6Mi15NU83PhMdG8dK6QSRoztLIpw8L2n+Qp8J+q3UOisceU1ISyydP5KJGh7BaqOfrwwtvv+dQbwEgZGMw2kNXMWoERQbUwKecT8aJpYRTP8OWEZAcz5aqvRgZuw9XWYYdPZZRwM3DobYpHk/ijEl0Wj2c69Y/8HdqwbquM3DS5Y0OGEeLQ96RPUW+4J9ftjB93Fgu6pzwSoxncO/etHvvA4cLw9nF/6I/dJVEvRa/d+vdXxjir8PqV2Btf/ANYFebBXwUewidtTDrXvpOCYPiLp7OrmzvNZ8qbh0ITd5Fi2X9uZ0Yn9tm5QjKc1A8Eozx8ayYMplQnR5hsVC/oC8vvvu+w+uxmq2c+fovvK8lEuOmp+L7ddG7Z9AdJCX8uw62fgDJCdB8DEdLteG1Xf0RCH58YSm1i/s73EbFk8HbW+ew+/oCXKz+rO60kLK+mXRZ5jBqnYPisePElk1s3r8fk6s7Xgmx9BowiGKVKju8HlOiiXNfHMM7wURMYVcqv10XrS4D5zjuGmx5D4I2Q8n60GkuwRofBmzoBcLEzGcXKmFQ3JdZbYcx40BRfrjwGZ3X92RBy29pVLpCbpvlMJQ4KHIMY3w8y6dM4pLOCaF3oqGPF23HjcuRuhJvJhE66zjeJivx5X2o0r9a+oOFUsI/a+CXD8GUBC0nwlNvcsuQSPeVvTBrohhT9ysV21+RJT54pit+XkX47NiHDNj1KhMbf0mnqo1z2yyHkKUxByFEGyHEWSHEBSHER/dJ97IQQgoh6tuP/YUQSUKIv+2f+Y4yXJG3ObE5kM8nTuCS3gWvhHje6NuXtu+8lyN1xYTGcOWLY3gkWzA2LEblARksbouLhJW94OeBULACvPEbNBlOosVMh9WDMWgu8XrFsfSs1TRH7FQ8mfSs1ZSvmy1CSCc+/nMoc/7YlNsmOYRMPQchhBb4BmiJbU/oI0KIQCnl6VTpPIHhwJ+pigiWUqodT/IJxrhYlk2ZzGW9M0Kno5GvNy/kkLcAcP3v68SuOIsTEs0LAZRrViptIinh5Gqbt2A2QKvJ0HgIaLRYLFY6rXyHWPEPLxYbzrtNOueYrYonl2Zlq7Pacxm9Ng1iftDHhMVGMrXVwNw266HISrdSQ+CClDIEQAixEugInE6VbiIwHRjhUAsVjw3/bN1M4L79mFzd8E6IpfegNyhSoWKO1Xdlz2XM20NBCDx6VqFwrcJpE8VGwOZ34dwvUKoxdPwGCpW/e7nnuk+IsP5Gfa+eTGv9eP8xK3KXyoX9+KXbCjqvfYMtEbO5ui6SHzqPyVNrIbJDVqwuCaTcgTvMfu4uQog6QCkp5eZ08gcIIf4SQuwTQjyTXgVCiEFCiKNCiKM3btzIqu2KPMTub+ex7vc/Met0NC7ow7szZuaoMFxYew65PZRkjYYiQ2qlFQYp4e8VMLcRhPwKrafAa1vvEYahm77kTFIgZZ1asahjhr2lCkWWKerhza7eSyipfZa/41fz4orhJJqMuW3WA5EVcRDpnLs7/1UIoQG+BNKblxgBlJZS1gHeA5YLIbzSFCblAillfSll/cKF02n9KfI066ZM4kB4JPpkIwO6d6fNW+/kWF1Wq5Uz357E5eg14p20lP6gHl5lUv2kYsNheXfY8AYUqQpDDsJTQ0Hz31qK8XuWciBqMYVEfdZ0nfbYtu4UeQ83vTNbe31NbY+uhJn30XJZP27Ex+a2WdkmK91KYUDKjlw/IDzFsSdQHfhVCAFQDAgUQnSQUh4FjABSymNCiGCgIqAWMjwBWCwWfvhkNJedXHFNjGfQO+9SoKRfztVnthL05XG8byUR46Gn0vv10bmm+AlLCX8vh22jwJIMbaZBw0GQ6sX/7eFfWHP5c9xlRTb0/CbbK1xNJhNhYWEYDAZH3JbiCWV01a7cTmpDkiWOf878RSHXgugdsNjTxcUFPz8/9PqcjQyclb+KI0AFIUQAcBXoAfS6c1FKGQMUunMshPgVGCGlPCqEKAxESSktQoiyQAUgxIH2K3IJU1IS88eO4Za7Fz7xsQwZNwFnj5xbSWxKSObc58fwTjITW8ydKsNqo0m5hiHmKmx6Gy7shDJNoMPXULBcmnI2nv6Tr099gpMszvqXFz7QLl9hYWF4enri7++PvUGkUGTItfhobhrCQWrx8yqNp7PrA5clpeTWrVuEhYUREJCz+0tkKg5SSrMQYhiwHdACi6WUp4QQE4CjUsrA+2R/FpgghDADFuANKWWUIwxX5B5x168xf8Y0Ety9KGlMpP+0GQ4Pf5GShOsJXJ79N14mCwlVClK1X7X/LkoJfy2F7aPBaoYXZkCDAWm8BYCDl87w8R/voJHuLGu/kBJeme8XnR4Gg0EJgyLLFPXwwUmrIzzhCpfjQilu8cPXzfOByhJCULBgQR7F2GyW/Gkp5VZga6pzYzNI2yzF93XAuoewT5HHiDxzmkVLvsfk5klVnaDbuOk5Wt/t89HcWPwPblaJuUlJKnVI4Q3EhEHgcAjeDf7P2LwF3/RbU0E3whi6awgIyTfN51GlyMN1fylhUGSHAq4e6DUBXIq9RETiFZItxSnmWeCBynpUvz21QlqRZc79updVO7ZjcXbh6cKFaDXsrRytL/JoJAlrzqNDom9fjhL/s0+SkxKO/wjbx4C0QtvPof7r6XoLAJFxt+kVOBCLJpbxDefwTEDVHLVboUgPD2cXyhUoS0h0KLeM4ZisJkp5F8ltszJETdFQZIkjq1eyYtcurBotL1avluPCcGl7KIY157AK8O5b7T9hiL4CP3WGTcOhRG0YeggaDsxQGOKMSXRaO4hkbThDq06gS7Wnc9Tuxw0P+zhReHg4L7/8crppmjVrRl4IhrlhwwZOn069vMqx5PS9uuj0VCxQFi2uxJpuEBx1lbwW/PQOShwUmbJ9zmy2/HsKjcVCzxYtaNCtR47Wd35FEGLPZQxaDcXeqk3BqgVt3sLR72HuUxB2BF6cCa8GQgH/DMsxWSx0XDWMBE0QL5d+n6GN2uWo3Y8zJUqUYO3atbltxn15FOLwKNBptVT09cdJeGKwRnM+6jIWqzW3zUqDEgfFfVk9aTy/37iFk9HAwFdfpWKz53KsLqvVyum5J3A9cYN4Fx3+IxvgWdIToi/DT51g8ztQsi4MOQQNMu5GulNW19UjuSEP08S33xO3rePSpUtp2LAhtWvXZvDgwVgsFuA/TwBg7dq19OvXD4Br167RuXNnatWqRa1atTh06NA95YWGhlK9um0XvqSkJHr06EHNmjXp3r07SUlJd9Pt2LGDp556irp169K1a1fi4217GUyYMIEGDRpQvXp1Bg0adLc13KxZM0aOHEnDhg2pWLEiBw4cyPTeFi1aRMWKFWnWrBkDBw5k2LBhHDp0iMDAQD744ANq165NcHDw3fQxMTH4+/tjtb9gExMTKVWqFCaTidmzZ1O1alVq1qxJjx5pGzX3u9eMnmW/fv144403eOaZZ6hYsSKbN6e39vf+aDQayvuWwl3ri0nGcy7qIskWc7bLyUnUmIMiXSwWC4vHjOKqixvuiXG88cFIPIvkXLx6c7KZszOP4x1tJMbLmUoj6qLTa+HIIthpn/vQ7iuo1w+yMCA3MHAGwcnbqejyIvPbO37fiDuM33SK0+GOXeBUtYQXn7avluH1M2fOsGrVKg4ePIher2fo0KEsW7aMV1/NWACHDx9O06ZNWb9+PRaL5e5LPT3mzZuHm5sbJ0+e5OTJk9StWxeAmzdvMmnSJHbt2oW7uzvTpk1j5syZjB07lmHDhjF2rO3/6ZVXXmHz5s20b98eALPZzOHDh9m6dSvjx49n165dhIeHM2DAALZuvWeeC+Hh4UycOJHjx4/j6elJ8+bNqVWrFk8//TQdOnSgXbt2abq/vL29qVWrFvv27eO5555j06ZNtG7dGr1ez9SpU7l48SLOzs5ER0dn+V4zIzQ0lH379hEcHMxzzz3HhQsXcHFxyVLeOwgh8C9QnKuxeqKTr3EhKoQAnzK46p2zVU5OoTwHRRqMcbF8PfIDrrq4UTAhlncmTM5RYTDEGDk7+Qje0UZi/Typ8lF9dAlh8GMH254Lfg1g6O9Q/7UsCcPonYs4HLOUYpqnWfXy5ByzO7fYvXs3x44do0GDBtSuXZvdu3cTEnL/5UN79uxhyJAhAGi1Wry9vTNMu3//fvr06QNAzZo1qVmzJgB//PEHp0+fpkmTJtSuXZsffviBS5cuAbB3714aNWpEjRo12LNnD6dOnbpb3ksvvQRAvXr1CA0NBWzdWKmFAeDw4cM0bdoUX19f9Ho9Xbt2zdIz6d69O6tWrQJg5cqVdO/e/a79vXv3ZunSpejSWeyY0b1mRrdu3dBoNFSoUIGyZcsSFBSUpXzpUdKrEEVc/ZDCTEj0RWINiQ9cliNRnoPiHm5fDWPBV1+S5OFFGZOBV6fm7BqG+PAEwub8hZfFSmLNwlTtWQmOLoKdn4LQQPtZULdvlkQBYNahjQRenY2XrMaGHrPR5aDtwH1b+DmFlJK+ffsyZcqUNNdSTnN8mBXc6U2XlFLSsmVLVqxYcc95g8HA0KFDOXr0KKVKlWLcuHH31O3sbGsJa7VazOb7d5086OBshw4dGDVqFFFRURw7dozmzZsDsGXLFvbv309gYCATJ07k1KlTaUQio6mh93uWqfM87PTSwu7e6LU6rsZf5kp8KEUtfhRyTxNp6JGiPAfFXa78/RdzZ88myc2DGs46Xps8NUeFISooiojZx3GxWLE2K0XFtq42b2HrCCjdyOYtZLEbCWDVyf0sPDseZ2spNnZdgLtz3nDPHc3zzz/P2rVruX79OgBRUVF3W/BFixblzJkzWK1W1q9ff0+eefPmAbYuw9jYjLvCnn32WZYtWwbAv//+y8mTJwFo3LgxBw8e5MKFC4Ctb//cuXN3X5yFChUiPj7+oQa2GzZsyL59+7h9+zZms5l16/5bJuXp6UlcXFy6+Tw8PGjYsCFvv/027dq1Q6vVYrVauXLlCs899xzTp08nOjo6TXdaRvcKGT9LgDVr1mC1WgkODiYkJIRKlSo98D3fwcfFHX+vAJBariVdISLu1kOX+TAocVAAcHrHNr5fvRqTkzPPlihOl1Ef52h9Eb+HE/39KbQSnDuXp0zBXTD3aYg4YVvM1udn8Elnb4YM2Bv8D5OOfoDW6sPKDgsp7JG7ra6cpGrVqkyaNIlWrVpRs2ZNWrZsSUREBABTp06lXbt2NG/enOLFi9/NM2vWLPbu3UuNGjWoV6/ePd0+qRkyZAjx8fHUrFmT6dOn07BhQwAKFy7MkiVL6NmzJzVr1qRx48YEBQXh4+PDwIEDqVGjBp06daJBgwaZ3kN4eDht27ZNc75kyZKMHj2aRo0a0aJFC6pWrXq3C6xHjx7MmDGDOnXq3DMgfYfu3buzdOnSu11KFouFPn36UKNGDerUqcO7776Lj49Plu71fs8SoFKlSjRt2pQXXniB+fPnZ3u8ISPcnVwoX6AsGpyJMkZyKfqaQ8p9EERem2Nbv359mRfmVOcnDi39kZ1nz4HVSvsG9anb6aUcre/i5hA0B8IwCkGhrr74nvwQLv0G5VvYupG8s7d6+WRkKH22vIIUFr5r8UOO7+N75swZqlSpkqN15Gfi4+Px8PDAbDbTuXNn+vfvT+fOeWcTpn79+qU7MO5IzFYLwVGXMZOIs8abcgVK3tN1ld5vUAhxTEpZ31E2qDGHfM6Wr2ZyJCoardlEr/YdKPd0kxyt79yPp3E9dZN4nYbSTc/jvm00aPS2TXhq985yF9IdrkTfpO+WQVg1SUx7ev4TtcF7fmXcuHHs2rULg8FAq1at6NSpU26b9MjRabRUKOhPyO2rGK0xnLtlopxvaXSanB1Du8eGR1aTIs+xbNwnnJcanA2JvD5wUI5uzmO1Wgn6+gReEfHEOGuoUPJLnA7tggqtbFNUvUtmXkgqopMS6PLzQEya67xfcwYvVnJYo0mRi3z++ee5bcJ9WbJkySOpRyME5X39uBStJ958k/NRFynrUwZnXc6G6r6DEod8iMVkYuHHo4h09cAjMZYhH43BvWDBHKvPbDBzduYxvGOTiXFJorKmP9ooDXSaB7V6ZttbADCYkumwegiJmmB6B4zhtXotc8ByhSL3KeNTlIg4HVHGSC7cvoi/d+lHUq8Sh3xGYkwM8yeNJ9bdiyKJcQyc9Bl6Z8cMpqWHIdpAyMzjeCdbiHMKpor1bTQVWkP7r8CrxAOVabVaeWn1CG7zF88XGcyopt0dbLVCkbco7lkQvUbPtaQwQmNDMT6CrUfVbKV8xK2LF5k9eSKx7l6Us5gYPGV6jgpD3JU4Lk0/iofRjIEdVHEbi6bLt9Br1QMLA0Df9ZO4Yt5LDfcuzGo7zIEWKxR5l0LuXpTy8AfgtjGKBUd+ydH6lOeQT7h09Ag/rV2L2dWNOu6udPxgXI7Wd/Ofm0QvPY0zgFhC+ZqJ0P4weBZ7qHLf+2Uuf8evwU/XjKUvpbuliELxxOLl4kZZbQDXiGT2qY+4Gnud8c/3zZG6suQ5CCHaCCHOCiEuCCE+uk+6l4UQUghRP8W5UfZ8Z4UQrR1htCJ7nNiyiR/Wr8es1/N8gD8dPxiZo/Vd3XeZuKWnEdKKq8scSvdqAz1XPLQwTNu/mh3X5uNDbdZ3+wLNfQLvKbKGCtmdfebPn8+PP/6Ya/W76p0p4lYYD1mRn8M+5/UN0+4GHXQkmXoOQggt8A3QEggDjgghAqWUp1Ol8wSGA3+mOFcV257T1YASwC4hREUppcVxt6C4H/u//449IaFogC5PNaZG25wNWx2y4gjav5MwWU0UqbAZn55fg+fDx2Vacnw3PwV/hqsMILDHPFz0Tg6wVnGHxyVkd7t27ahaNXc3a3rjjTdytX4ArUbD7l5L6bj6LQ7HLKXjqkiH15GVpldD4IKUMkRKmQysBDqmk24iMB1IGYSkI7BSSmmUUl4ELtjLUzwCNs6Yxp7Qy+hMJvp27pyjwmBJNnFm6ib0fyeRaE3Cr1MMPgNnOkQYtp49xhcnPkJnLcy6l76jgJtH5pmecFTIbseE7F6yZAnDhv03btWuXTt+/fXXu89yzJgx1KpVi8aNG3Ptmm218rhx4/LEdFt3Z2e29ZpHVbeOhCbvcnj5WRlzKAlcSXEcBjRKmUAIUQcoJaXcLIQYkSrvH6nyppnQLoQYBAwCKF360UzTepKxWCwsHfcJF7VOuCQmMHDYWxQs459j9d08eIIbG2/gqfEhRhNLxffroC/y4APOKTkadoGRB99C4MKStgsp7VPYIeU6jF8+gsh/HFtmsRrwwtQML6uQ3Y4L2X0/EhISaNy4MZMnT+bDDz9k4cKFfPxxzoaVyS46rZZVXSfxzi/F+RfHTs7IiueQ3iT0uzE3hBAa4EsgvaD5981794SUC6SU9aWU9QsXzmN//I8ZJqOB+aM+5KLWCa/4WIZ/PDbHhMGSlMyZqZtICIzBGT3J5eOpMqmNw4Qh+FYkA7YPBszMfHYOtYv7O6Tcxx0VsjstDxqy+344OTnRrl27NLbnRb564U2Hl5mVpxUGpIyA5geEpzj2BKoDv9pjfxQDAoUQHbKQV+FAEm7dYt7UycS7e1EsKYGBU6ah1efMasobv/3NrcCbdm8hjjKvV8ejvOO8vpsJcXTfOBCzJooxdb+iRflaDivbodynhZ9TqJDdaXnQkN06ne6ewdyUduv1+rvPISu2P2lkxXM4AlQQQgQIIZywDTAH3rkopYyRUhaSUvpLKf2xdSN1kFIetafrIYRwFkIEABWAww6/CwXXzp9j9rQpxLt5UhErb0ybkSPCYPMWAkncFIsTepIrJFBtSluHCkOiyUiHNYMwaC7xesWx9KzV1GFlPwmokN1pedCQ3f7+/vz999930x4+rF5Pd8hUHKSUZmAYsB04A6yWUp4SQkywewf3y3sKWA2cBrYBb6qZSo7nwsEDLFi8CKOLGw0LeNNr3IQcqefG/uOc/3Q/ntEFiNfGU3RIOcq+3sahdZgtFjqufJs48S8vFn+Ld5vknWiceQUVsttxIbubNGlCQEAANWrUYMSIEVneJjQ/oEJ2P+YcW7+OzceOA4JWVSrxVO9XHF6HOcnA+a924B7tg1lKqJxE2f6OFYU7dF09hqCkQOp79eT7zqNzpI7/t3feYVEdbR++h94ECxqxl2AUDGBEECtGY4k1iSV2NLao0bx5NTHxNRpjisYU/WyxxIIdNWosUaNiLLGgAQsWBLEiKiJIX3bn+wPYoBQRKQuZ+7r22t1zpp6B8+xzZub5vSgqZHfhYughuw0BFbJbkSv7Fy3g8O0IjHQ6erbxxumNgt9jeM//NA93RVPGqBwxxo+pNdwF69rPp7eQV0b/9iOXErdTx6w9y7rnuNdSUcpRIbsNA2UcSiibv57BuWQNpinJ+PTvT1WXgp2wTU1MJOTHfVjHlMUUUzSvJOA8JOtjgILiiwO+HH74C/bCHb9eM9Xu538xhrCHQKGMQ4lDq9Wycspn3DCzxDIhjpEf/peyVQtm6WgG9w4G8HB3zD/ewgg3rGsVbB2ZWXRyF343ZmMt67G173zMnnPJoUKhKHjUf2EJQpOYyMLP/8dD6zKUi49l1NTpmNsU3G7h1PhErszZh01MWUwxQVM/EWefwvMWAH69cJx5F6ZgcUin9AAAIABJREFUJqvwa88l2FlYFWp9CoUibyjjUEKIjYhg0Q/fkWBtS7XkBIZ8+x3GxgUnGRh54BTRv8dia1SOGONYao96DasaDs/O+AIcvX6Rz0/8ByNpw5quS6hiW75Q61MoFHlHGYcSwJ0L51m+aiUaqzI4mQh6T5tVYGWnxiVwZc4f2MSmewtOiTgP6lxg5efEpXu3GP3H+yAk819fSINKhTPJrVAo8ocyDgbOpQMH8PtjL1pzC5pVtKf92A8KrOzI/ad4tCfdWzCJpfbIwvcWACIeR9Pvt+FojR4z3WMeLWsXb5RNhUKRFbUkxIA5uWEdGw7sR2dkTOdXGxaYYUh9nEDwjO0k703EBGNSnZNx/qpzkRiGx8mJvLVpBCnGdxjj/CVvO3sVep2KnFF6DkXH0xFgDR3lORgoe+bN5a979zHWanm3U0ccW3kXSLl3950gZl/cP97C6MZYVX0xEZ68otFq6b5hLPFGl+hZfSLvexTuZLci7yg9B8XTKONggGz4choXUyVmyUkM9RlC5QYv/s+Q+jieKz/txyauLMYYo301Gef+hT+3kIFOp6Pnxo+5L0/SovwQpr2ec3jpksTMkzO59PBSgZZZv3x9PvHIXa1v9erVzJ07l5SUFDw9PVmwYAHGxsbY2Njo4wdt2rSJHTt2sGLFCiIjIxk1apQ+euvChQtp1qyZvrzw8HC6dOnC+fPnSUxMZMiQIQQHB9OgQYMseg5Tp04lOTmZunXrsnz5cmxsbJg+fTq//fYbiYmJNGvWjJ9//hkhBN7e3nh6enLw4EEePXrEsmXLaNmyZa59W7ZsGTNnzqRKlSo4Ojpibm5Ov3792L59O4cOHWLGjBls3ryZunXrAml6Dq6uroSFhWFkZERCQgKvvPIKYWFhLFy4kEWLFmFiYoKTkxPr169/oq7w8HAGDhxIfHw8APPmzaNZs2b06dOHwYMH60N8+Pj40LVrVzp16oSPjw+XLl2iQYMGhIeHM3/+fNzds9+YvHz5cr755hscHByoV6+ePgihj4/PE+HHM8bN39+fzz//nAoVKnD58mVatWrFggULimXfj3qsZEBotVqWTPqYi1qwTnjMB/+dUCCG4e7e44R+eQLb+HI8No2l6ocNqNm/XQG0OO8M2z6TsJS9vGLZhYVdPyrSuksbmfUcAgMDMTY21gfKy4kMPYegoCDOnDmDs7Nzjmkz6zlMnjyZ06dPA0/qOZw5cwZ3d3d++OEHAMaOHcupU6f0xmXHjh368jL0HH766Se++OILIOfYShl6DsePH2ffvn1cupRmeDP0HL777jsCAwP1hgGe1HMAsug5/P3335w9e5ZFixZlqa9SpUrs27ePM2fOsGHDBsaNGwekxXHKCAGekpLC/v37efPNN1mwYAHlypXj7NmzTJkyRX9tsiMiIoKpU6dy9OhR9u3bl+dHYidPnuT777/n3LlzhIaGsmXLljzlK2iU52AgJD+OZeH0aTyytqVC/GNGTf8KU0vLFypT8ziekB/3YxNfFmOM0Lqm4Ny3cGVCs+PTvUs5FbOWykbNWP/OjCKvvzB51i/8wiCzngOkKbdVqlQp1zwHDhzQ6x7nRc8h4yaZk54DpN00vbzS5owOHjzIrFmzSEhI4OHDhzg7O+vFfvKr5wDQq1cvrly58sxrkqHn0KZNG9avX8/o0aP17e/fvz89evTINgyHRqNh7NixeiObUVenTp0YN24cycnJ/P7777Rq1QpLS0uOHDnC+PHjAWjYsKH+2mTHiRMn8Pb2JkOjpk+fPnnqi4eHB3Xq1AGgb9++HDlyJMf5oMJEGQcDIPr2LRb/9COJ1rbU1CQx6NtZL7yHIWL3X8QeTEybWzCNoc5YTywrF72Q0k/HtvHbnf/DVjqz9d25mBTg3ox/K0rPISv51XP48ccfeemllwgKCkKn02FhYQGAhYUF3t7e7Nmzhw0bNtC3b998tS+76whP6khIKUlJSckxT05lFDbqsVIxczPwbxbMnUuilTUuFqYM+erbFzIMmtg4gqdvJ9Vf84+3MKNLsRiGdUF/svTyF1joqrOt12Ks028SihdD6TlkJb96DjExMTg4OGBkZISvr69eixvSHi0tX76cw4cP06FDWlDLFi1asHHjRgCCg4M5dy5niVhPT0/8/f2JiopCo9Hg5+enP1erVi39I6lt27ah0Wj0506ePMm1a9fQ6XRs2LCBFi1a5PXyFSjKOBQjwXt/Z/nGjWjMzGlVpQpvT5r8QuVF7DrGtRknsU0oR6xZLFU/cqZm37YF1NrnY3/oWb4+PQFjXVk2dF9KRRvbYmlHaUTpORScnsPo0aNZuXIlTZs25cqVK1hbW+vPtW/fnj///JN27dphZmamT3///n1cXFyYOXMmLi4uOT6ic3BwYNq0aXh5edGuXbsntCKGDx/OoUOH8PDw4MSJE0/U6+XlxaRJk2jYsCG1a9cutnDlSs+hmDi2ehX7Ll8BnY6uTdx5rcfb+S5L8+gxIXMOUiahLMlSh1ljHTX6vF6ArX0+giLCGbh7IBItv7RfRZNqLxdbWwoDpedQuBiynoNWq0Wj0WBhYUFoaCht27blypUreuPxovj7+zN79uwnJvSzw2D0HIQQHYE5gDGwVEr57VPnRwFjAC0QB4yQUgYLIWqRph53OT3pcSnlqIJpesll54/fcyo6BuNUDf27daeOV7NnZ8qBiJ3HeHwoKW1uwSyGumObYvGSfQG29vm4+egBPrtGoDNKZGazRaXOMCgKH0PWc0hISKBNmzZoNBqklCxcuLDADIOh8UzPQQhhDFwB3gBukaYp3VdKGZwpja2UMjb9czdgtJSyY7px2CGlbJjXBpV2z2HNtCmESCPME+N5b/gIKjnWy1c5mkexXJnjj21CWZKkDgt3SfXebQq4tc9HdEI87df3J1GE81+X7xjS+I1ibU9hoTwHxdN4enqSnJz8xDFfX19effXVQqnPUDwHD+CqlDIsvQHrge6k6UIDkGEY0rEGDOtZlQGg1WhY/L9PibS0oUxCLKMmTca6QoV8lXXnt6PEHU7GzqgcMeaPqDu2GRaV8ldWQaDRaVh3fgdzTy8hyegWA+pMLrWGQaHIjhMnThR3EwqcvBiHqsDNTN9vAZ5PJxJCjAE+AsyAzA+8awsh/gZigf9JKQ9nk3cEMAKgRo0aeW58SSEhOppFX31JrI0tlRLiGD7ja0zNLZ67nJToGELm/Iltoh0CI6S7FueeXQuhxXkjJjkGv8ub+OXcah6nPoBUe/o6fsqkVn2KrU0KhaJgyItxyG6RbRbPQEo5H5gvhOgH/A8YDEQANaSUUUKIxsBWIYTzU54GUsrFwGJIe6z0nH0waKKuXWPJwvkk2dhSV6eh3zcz87VU9fb2w8QfScXOqGyxewvXY6+zOng1v17dSrI2idT4urxi0Y95PftRrZz1swtQKBQGT16Mwy2geqbv1YA7uaRfDywEkFImA8npn08LIUKBekDpnVTIxPVTJ/DdvIVUSytes7Gi24SPn7uMlKhHhMw9jG2SXZqVbqLF+Z2i9xaklAREBrAqeBWHbh5CYExqrCtGsa34osMb9Gpcrdg26ygUioInL8bhFOAohKgN3AbeBfplTiCEcJRShqR/7QyEpB+vCDyUUmqFEHUARyCsoBpvyATt2M62v44jTU1pW7sWLX2GPncZt7ceJv5Yurdg8Yi6HzTHwr5o1dI0Wg2/h/+Ob7AvFx9epIypHeU1b3I93BXvunX5etCrONi9WJgPhUJheDxzE5yUMhUYC+whbVnqRinlBSHE9PSVSQBjhRAXhBCBpM07DE4/3go4K4QIAjYBo6SUDwu8FwbGoV+W8uuJUwC87dX0uQ1DSlQ0F6b+hjye/kzPU+L8RdciNQyPkh6x5OwSOmzuwGdHPiNZm0zbimN4eOljom614bu3WvKLTxNlGEo4Ss/BsMhtHIqaPO1zkFLuAnY9dezzTJ/H55BvM7A5u3OllW3fzeTvuHhMNMkM7N2bmq8938qyW1v+JPG4Vj+34DiuBWYVyhVSa7NyLeYaq4NXsz10O0naJJpVacbohp+y7pAlW6/H8Hr9Snz91qtUtnv+CXWF4aL0HAwDQxoHFXivgNBqtayeOoVrJmZYJMQzfOwHVKhZK8/5kx48JHTuUexSyiIB0RScexTN3IKUkhN3T7DqwioO3z6MmZEZXep2od8r/TkcbMLktZcxN4nn+16uvP1aVTW3kIm7X39N8sWC1XMwb1Cfyp99lmsapedQMvQcchoPHx8fbG1tCQgI4O7du8yaNYuePXs+MQ7FjYqtVABokpNYOOljrpmYYRsXy7j/ff5chuHWZn9uzgrCNtmOGItH1Pr0Nar2yP0fqCBI0abwa8iv9PytJ8P3DudC1AVGu45mb8+9DHKcyKcb7jNj50VavGzPvo9a846adDYIlJ5DydBzeBYREREcOXKEHTt2MGnSpHyXU1goz+EFiY+KYuE3XxFnY4tDYhzDvpmJsalpnvKmeQvHsEuxQ5KKUXOBc7fC9xYeJj1kw+UNrL+0nodJD3m57MtMbzadN+u8iYkwY9mRML7fewULU2N+7ONKDzflLeTEs37hFwZKzyErhqjn8Cx69OiBkZERTk5OREZG5rucwkIZhxcgMuQKvyxZTLJ1GeoJHf1mzs5z3pub/Ek6KbEVtsRYRuM4vjVm5XL+hy0IrkZfZfXF1fwW+hspuhRaVm3JQKeBNHVoihCCq/fimLgpgL9vPKJdg5f4+q2GVLJVcwuGhtJzyEpJ0HN4ejzMM4WwN7QAqKAeK+Wbq0cPs/iXZSRbWOFRzo5+U6fnKV/SvSguTNmBCDBGIjFuaYzztG6FZhiklBy9fZSR+0by1va32BG2g+4vd2db920saLcArype6CQsOhTKm3MPc+1BPHPedWPJoMbKMBgoSs8hK4ao5wA5j0dJQHkO+eD0r5vZcfoMGJvSoX49vPoNyFO+mxsPkhQg0rwFq2jqjffGtGzh6BwkpSaxM2wnvsG+hMaEYm9pzweNPqBXvV6Us/hn9VNI5GMmbDpL0M1HtHd6iRlvNaRSGWUUDJnMeg46nQ5TU1Pmz59PzZo19XoO1atXp2HDhvqb4Zw5cxgxYgTLli3D2NiYhQsX6h8JPc3777/PkCFDcHFxwc3NLVs9h4wgczNmzKBevXp6PYdatWrlWc9h2LBhWR4tZdZzqFKlShY9h+HDhzN37lw2bdr0xLwDpD1a6tWrF/7+/sA/eg4xMTFIKXPUc3jnnXfw8/OjTZs2WfQcBg0aRLdu3Z7Qcxg8eDAuLi40atQoVz0HIMfxKAkoPYfnZP+iBRy+HYGRNpVer79Og3bPDjCXFPmA0HnHsdPYkaDTUKa1BQ6d8x+mOzceJD5g/aX1bLy8kejkaOqXr89Ap4F0rNURM+N/QgunanUsPhzGT/tCsDY35ovuDenq4qDmFvKAispauPyb9RzyiqFEZVWks+nrLzmfnIppSjI+/ftT1cX1mXlubDhAymkjbIUtsVbROBaSt3D54WV8g33ZdW0XqbpUWldrzSDnQbi/5J7lhn8l8jET/YIIuhVDR+fKfNmjIRXLKAlPhWGg9BwMA2Uc8oBWq2XFlM+4aWaJVUIcIz78L2WrVsk1T+Ld+4TNP4Gdxg4dGky8zXDq1C3XPM+LTuo4cvsIq4JXcSLiBJYmlrzj+A4DnAZQ07ZmlvSpWh0//xnGnD9CsLEwYV6/RnR+VXkLCsNi9uy8L+woasqUKZPtbvGi1nMoCpRxeAbJcXH8/MXnPLS2pVx8LKOmTsc8PeRATlxfvx/NGeM0b8E6GscPX8fUNvc8z0NiaiK/hf6Gb7Av4bHhVLKqxIevfUjPej2xM8/++eflu4+Z4BfEudsxdH7VgS+6O2Nvo7wFhaIg+LfqOfxriY2IYNEPs0mwtqVaciJDvv0u13DbiXfuEbbgJHapdiShwcTbvEC9hXsJ91h3aR1+V/yISY7BuYIz37b8lva12mNqlP3eCo1Wx8+HQpmzPwRbC1Pm93uNzi4O2aZVKBSKDJRxyIE7F86zfNVKNFY2OJkY0XvazFzTX1+7H01g4XgLwVHB+Ab78nv472h1Wl6v8TqDnAbRqFKjXB8JXYyIZeKmIM7fjqWziwPTuzlTQXkLCoUiDyjjkA2XDhzA74+9aM0taPaSPe1Hf5Bj2sQ7kYQtCMAu1TbNW3jdHKcOL+4taHVaDt06hG+wLwGRAViZWNHnlT70r9+f6rbVc82r0epY6B/K/x0Iwc7SlIX9X6PTq8pbUCgUeUcZh6c4sX4tv5+/AEbGdH61IU165Sx5eX3NH2iCTCgjyhBbJpp649tiUubFlNASNAlsvbqVNRfXcOPxDRysHZjgPoG3HN/C1uzZq5yC78QywS+I4IhYurpW4YtuzpS3Lp2rKRT5JyAggFWrVjF37twiqS88PJxjx47Rr1+/ZyfOJytWrCAgIIB58+YVWh3/JpRxyMSeeXP4614Uxlot73bqiGMr72zTJdy+y7UFp7HT2pJECmbtLHF648W8hbvxd1l7aS2brmziccpjXCq68MFrH9CuRjtMjJ49TCmpOhb4X2XegauUtTJj0YDGdGxY+YXapCi9uLu75xhJtDAIDw9n7dq1hWocFAVLnoyDEKIjMAcwBpZKKb996vwoYAygBeKAEVLK4PRznwLvpZ8bJ6XcU3DNLzg2fDmNi6kSs+REhg59j8qv1M82XbjvH2jPZfIWPmyHiY1Vvus9/+A8q4JXsTd8LxJJuxrtGOg0ELdKbnku48KdGCb4neViRCzd3aowrasz5ZS3UCQc3niFBzcLdterfXUbWvaul2uaVatWMXv2bIQQuLi44Ovry/Xr1xk6dCj379+nYsWKLF++nBo1auDn58cXX3yhD7j3559/4u/vz+zZs9mxYwfTpk3jxo0bhIWFcePGDT788EN94L2cQoPnRGhoKP3790er1dKpUyd++OEH4uLimDRpEhcvXsTNzY3Bgwfzn//8R58np/DY9evXZ8iQIaSkpKDT6di8eTOOjo5P1Ld8+XK++eYbHBwcqFevnj5ekY+PD126dNEL52SEzvb39+fzzz+nQoUKXL58mVatWrFgwQKMjFQkoad55hURQhgD84FOgBPQVwjxtNrGWinlq1JKN2AW8EN6XifSZEWdgY7AgvTyDAatVsviSR9zUQvWCY/54L8TsjUMCTciuDB5JyYXzElFi/kbFjj9r1u+DINWp2Xf9X0M2j2Ivjv7cvjWYfo36M+ut3fxvff3eTYMKak6fth3he7zjnL/cTI/D2zMnHcbKcNQyrlw4QJfffUVBw4cICgoiDlz5gBpYbMHDRrE2bNn6d+/v/4GP336dPbs2UNQUBDbt2/PtsxLly6xZ88eTp48yRdffIFGo8k1NPiwYcOyXe8/fvx4xo8fz6lTp6hS5Z+9QN9++y0tW7YkMDDwCcMAOYfHXrRoEePHjycwMJCAgACqVav2RL6IiAimTp3K0aNH2bdvX55V4k6ePMn333/PuXPnCA0NZcuWLXnK928jL56DB3BVShkGIIRYD3QH9CMhpcwcxcsayIjJ0R1YL6VMBq4JIa6ml/dXAbT9hUl+HMuC6dOIsbalQvxjRk3/ClPLrLKX4av2oT1vRhlRhse20TiOz5+3EJcSx69Xf2XNxTXcjrtNVZuqfNLkE3q83AMbs+db2XT+dgwT/IK4dPcxbzWqytSuTpS1UkahqHnWL/zC4MCBA/Ts2RN7e3sAfXjrv/76S3+jGzhwIB9//DEAzZs3x8fHh969e+vDZz9N586dMTc3x9zcnEqVKhEZGZlraPClS5dmW85ff/3F1q1bAejXrx8TJkx4Zn9yCo/t5eXFV199xa1bt3j77bezeA0nTpzA29ubihUrAmkeSF7Ce3t4eFCnTh0A+vbty5EjRwxGmtOQyItxqArczPT9FuD5dCIhxBjS9KPNgNcz5T3+VN6q2eQdAYwAqFGjRl7a/cJE377F4p9+JNHallqaZAZ+OyuLu5xwI4Jri85gp0ubW7DpYEuN11s9d123426z9uJatoRsIU4TR6NKjZjgPoE21dtgbPR8jlRyqpZ5B66ywD+U8tZmLBnkzhtOLz13mxQlFyllnna1Z6RZtGgRJ06cYOfOnbi5uREYGJglbebw0RmhtXMLDV6Q5BQeu1+/fnh6erJz5046dOjA0qVL9aG4M8jpOpiYmKDT6YC065WSkpJjHhUhIHvy8qAtuyuXJVqflHK+lLIu8Anwv+fMu1hK6S6ldM/4FVCY3Az8m/n/N5dEK2tcLczw+eqbLIbh2sq9RMy7Qhltmrfw8udevPT6s6NNZibwXiD/9f8vb255kzUX19CyWkvWdV7Hqk6raFez3XMbhnO3Yuj2f0f5vwNX6e5WhX3/aaUMw7+Qtm3bsnHjRqKiooC0kN2QppaWIYO5Zs0aWrRoAaTNA3h6ejJ9+nTs7e25efNm9gVnU09OocFzomnTpvow25klOXMLtw3Zh8cOCwujTp06jBs3jm7duulDh2fg6emJv78/UVFRaDQa/Pz89Odq1aqlV2nbtm0bGo1Gf+7kyZNcu3YNnU7Hhg0b9NdJ8SR58RxuAZkX1lcD7uSSfj2wMJ95C53gvb+zyf8QOlNzWlerSpvhI584Hx9+h/DFgdjpyhBHMmU6lqVGm7x7C6m6VP648Qe+F3w5++AsZczKMNh5MP3q96Oydf5WDyWnapm7P4RFh8KwtzFj2WB32jZQRuHfirOzM5MnT6Z169YYGxvTqFEjVqxYwdy5cxk6dCjfffedfkIaYOLEiYSEhCClpG3btk9IauZGbqHBhw0bxqhRo7KsePrpp58YMGAA33//PZ07d9aHs3ZxccHExARXV1d8fHyyzDtkFx57w4YNrF69GlNTUypXrsznn3/+RB4HBwemTZuGl5cXDg4OvPbaa3o9huHDh9O9e3c8PDxo27btE6G4vby8mDRpEufOnaNVq1YGE/HV0HhmyG4hhAlwBWgL3AZOAf2klBcypXGUUoakf+4KTJVSugshnIG1pM0zVAH2A45SSi05UJghu4+uWskfISGg09HNowmNuj/5R3Ft+R50Fy0wFYJ4u0c4/ucNTLKZg8iO2JRYtlzZwtpLa4mIj6BGmRoMcBpA97rdsTLN/2qms7ceMcEviCuRcfRsXI0pnZ2ws8qbDKmicFAhu3MmISEBS0tLhBCsX7+edevWsW3btuJulp7Mq7RKMgYRsltKmSqEGAvsIW0p6y9SygtCiOlAgJRyOzBWCNEO0ADRwOD0vBeEEBtJm7xOBcbkZhgKk50/fs+p6BiMNRr6d+9OHa9/9BTir90ifMlZvbdg27kcNVrlzVu4+fgmay6u4deQX0lITcD9JXc+9fiUVtVaPfdjo8wkp2qZ80cIP/8ZRkUbc5b7NKFN/dx1ghWK4ub06dOMHTsWKSVly5bll19+Ke4mKfLJv0LsZ820KYRII8wT4xk+chT2dV/Wnwv75Xe4ZImJEMSXfYTjh8/2FqSUnLl3Bt9gXw7cOICxMKZT7U4McBqAU4WnV/k+P4E3HzHRL4iQe3H0dq/G5M5O2Fkqb8FQUJ6DorgxCM+hJKPVaFj8v0+JtLShTHwsoz6djHWFCgDEhd7k+rJz/8wtdC7/TG9Bo9OwN3wvvsG+XIi6gJ25HcNeHca79d+lktWL/6pP0mj56Y8QFv8Zyku2FqwY0gTvV5S3oFAoip5SaxwSoqNZ9NWXxNrYUikhjuFffY2puQU6nY7wX/bCFUtshA2Py0VTb3xHjC1z3iMQkxyD3xU/1l1ax72Ee9SyrcWUplPoWrcrliZ5m5N4FmduRDPRL4jQ+/G826Q6n3VugK2F8hYUCkXxUCqNQ9S1ayxZOJ8kG1vq6lLp981MjI2Nibt6g+vLzmMn07yFCl0rULNFzt5CeEw4qy+uZnvodhJTE/F08GSq11RaVG2BkSiY7fZJGi0/7rvCksNhVLa1YOVQD1rXK/zlvAqFQpEbpc44XD91At/NW0i1sKJxGSu6/vdjdFotYUt2Q4gVNsKGuPLROI7L3luQUnLq7il8g305dOsQJkYmdK7TmQENBvBK+VcKtK2nr0czcVMQYffj6etRg8/erE8Z5S0oFAoDoFQZh6Ad29n213GkqSlt69Smpc8QHl8J58byi9hJGx6TTMWu9tl6Cxqtht3hu/EN9uXSw0uUMy/HSNeR9HmlD/aW9gXaziSNlu/3XmbpkWtUsbPE9z0PWjoqb0FhGLz55pusXbuWsmXLFkl9P/30EyNGjMDKKv9LvguCYcOG8dFHH+Hk9OKLSkoDpWa10qFlSzgYfgMjXSpvtWiBc/uOXFu6FxFqhbEQJFZ4hOMHWb2F6KRo/K74sf7Seu4n3qeuXV0GOg2kc53OWJhYFFS39ASEP+TjTWcJexBPf88aTOqkvIWShlqtVLDUqlWLgIAAfawoxbNRq5XyyNZZ3xIYn4CJJpmBvXtT3saei5P3YocNj0miUo+XqOn1pLcQFhPG6uC0+YRkbTLNqzTny+Zf0qxKs0KJtZKYomX23sv8cjTNW1gzzJPmL6t/hpLOwRWLuXc9rEDLrFSzDm18RuSa5nlCdvv4+GBra0tAQAB3795l1qxZ9OzZk4iICPr06UNsbCypqaksXLiQli1b6m/WcXFxdOrUiRYtWnDs2DGqVq3Ktm3bsLS0JDQ0lDFjxnD//n2srKxYsmQJ9etnH+YeQKfTMXbsWA4dOkTt2rXR6XQMHTqUO3fucOfOHdq0aYO9vT0HDx7U59m9ezfLly9n48aNQNoGtu+//56tW7fy3nvvERAQgBCCoUOHZtlxnVvI7mnTpmFvb8/58+dp3Lgxq1evRgiBt7c3s2fPLlKdC0OmRBsHrVbL6qlTuGZihkVCPMPHfED07xe5fy0eG6yJs3+UthLJLO2XuZSS4xHHWRVmbpGOAAAW/ElEQVS8iiO3j2BmZEbXul0Z0GAAL5d7+Rm15Z9T4Q+Z6BdEeFQCA5vW5JNO9bExL9GXXlGMZITsPnr0KPb29vrYShkhuwcPHswvv/zCuHHj9BFSIyIiOHLkCJcuXaJbt2707NmTtWvX0qFDByZPnoxWqyUhISFLXSEhIaxbt44lS5bQu3dvNm/ezIABAxgxYgSLFi3C0dGREydOMHr0aA4cOMD27dsJCAhg+vTpT5SzZcsWwsPDOXfuHPfu3aNBgwYMHTqUcePG8cMPP3Dw4MEsnsMbb7zByJEjiY+Px9ramg0bNtCnTx8CAwO5ffs258+fB+DRo0fPdf3+/vtvLly4QJUqVWjevDlHjx5V8ZWyocTeoTTJSfw8ZTIPrMpgFxdL/35DuLvo4j/ewluVqdk0zVtI1iazK2wXvhd9CYkOoYJFBca4jaH3K70pb1G+0NqYmKJl1p5LrDgWTtWylqwd7kmzuspbKE086xd+YfC8IbsBevTogZGREU5OTkRGRgLQpEkThg4dikajoUePHri5ZdURqV27tv5448aNCQ8PJy4ujmPHjtGrVy99uuTkZAC6detGt25ZVRGPHDlCr169MDIyonLlyrRp0+aZ/TQxMaFjx4789ttv9OzZk507dzJr1ixSU1MJCwvjgw8+oHPnzrRv3z5P1y0DDw8PvTaEm5sb4eHhyjhkQ4k0DnH377No1jfEWdvikBhPm9qvE7/hYZq3UPER9caleQtRiVFsvLyR9ZfX8zDpIfXK1ePL5l/yZu03MTMuXO2DE2FRfLz5LNejEhjkVZNPOtbHWnkLigLgeUN2w5MhuTPmGVu1asWff/7Jzp07GThwIBMnTmTQoEFPlPF0KO/ExER0Oh1ly5bNNvR3bm3OD3369GH+/PmUL1+eJk2aUKZMGQCCgoLYs2cP8+fPZ+PGjVnCdOQWsju78OSKrJQ4bbzIkCvM/W4mcVZlqJcKLURbrG7akUQKZd4uT/0JXQmLD2fqsam039SeBUELaGjfkCXtl7Cp6yZ6vNyjUA1DQkoq07ZfoM/i40gJ64Y3ZXr3hsowKAqM5w3ZnRPXr1+nUqVKDB8+nPfee48zZ87kqX5bW1tq166tD5EtpSQoKCjXPC1atGDz5s3odDoiIyPx9/fXn8stnLe3tzdnzpxhyZIl9OnTB4AHDx6g0+l45513+PLLL7Ntd24huxV5o0Tdsa4ePcy6HTvRWljhllIWV91rGCGIr/QIx7EdOP7gJL77RnLszjEsjC3o8XIP+jv1p45dnSJp3/GwKD7edJYbDxPwaVaLjzu+gpVZibrEihLA84bszgl/f3++++47TE1NsbGxYdWqVXluw5o1a3j//feZMWMGGo2Gd999F1dX1xznHN555x32799Pw4YNqVevHp6envpw3iNGjKBTp044ODg8MSENab/su3TpwooVK1i5ciUAt2/fZsiQIXrPIDsxotxCdivyRolZynp68yZ2/P03IGiV4oSjUXViZSIV3q7I4XKh+Ab7EhYTRkXLivSt35de9XpR1qJo1mnHJ6cy6/dLrPzrOjUrWDHzHRea1qlQJHUrih61lDV/xMXFYWNjQ1RUFB4eHhw9epTKlfOncfJvRy1lTWf/ogUcvh2BGSa8qfGkrLDhYcUoDrW+gd/VGURfiqZB+QZ83eJrOtbqiKlx0e0bOBb6gE82n+VWdCJDmtdiYgflLSgU2dGlSxcePXpESkoKU6ZMUYbBwDH4u9imr7/kfHIqttKKLqlNSEGH32uHWJfyK6kXUvGu7s1Ap4G4v+RepFqw8cmpfLv7Er7Hr1OrghUbRnjhUbvwVj4pFCWdzPMMCsPHYI2DVqtlxZTPuGlmSRVdedpoXuVkmSvMrLoQi1RzetbryYAGA6hhW6PI23bs6gM+3nyW248Sea9FbSa0fwVLs/wL+ygUCoWhkSfjIIToCMwhTQluqZTy26fOfwQMI03t7T4wVEp5Pf2cFjiXnvSGlDLrIuinSI6LY/G06UTZWFEv1YG6KdWZWOP/iHvpMR82GM87ju9gZ26X504WFHHJqXyz6yJrTtygtr01fiO9cK+lvAWFQlH6eKZxEEIYA/OBN4BbwCkhxHYpZXCmZH8D7lLKBCHE+8AsoE/6uUQpZdbdNTmg1WhY8OVsYmyscNXU5IZpFDs8VjHy1RG0q9kOU6PiiUN0JCRtbuFOTCLDW9bmozeUt6BQKEovefEcPICrUsowACHEeqA7abrQAEgpM68/Ow4MyG+DHtyLItZC4JJYlfNuV+nStifjKv63SOcTMvM4ScPXuy6x7uQN6thbs2mUF41rKm9BoVCUbvJiHKoCNzN9vwV45pL+PWB3pu8WQogA0h45fSul3Pp0BiHECGAEQBWHKrycaEWTSe14u3ytPDSv8Pjzyn0mbT7L3dgkRraqw3/eqIeFqfIWFApF6ScvO6Sz+8me7eYIIcQAwB34LtPhGulrb/sBPwkh6mYpTMrFUkp3KaV7edsy9J81kerFaBhikzRM2nyWQb+cxNLMmE3vN+PTNxsow6AoNQQEBDBu3Lgiqy88PJy1a9cWWX2FgY2NTXE3oUjJi+dwC6ie6Xs14M7TiYQQ7YDJQGspZXLGcSnlnfT3MCGEP9AICM2pMvNiHoBD6d5CZGwSo1rX5cN2jsooKHLk0W+hpNyJL9AyzapYU7Zrlt9QBYq7u3uRhqbOMA79+vUrsjoVL0ZePIdTgKMQorYQwgx4F9ieOYEQohHwM9BNSnkv0/FyQgjz9M/2QHMyzVUYErFJGj7ZdJbBv5zE2tyELaObM6lTfWUYFAbJqlWrcHFxwdXVlYEDBwJpsZLatm2Li4sLbdu25caNGwD4+fnRsGFDXF1dadUqLVKxv78/Xbp0AWDatGkMHToUb29v6tSpw9y5c/X1rF69Gg8PD9zc3Bg5ciRarTbXdoWGhtK0aVOaNGnC559/rv+1PWnSJA4fPoybmxs//vjjE3n69OnDrl279N99fHzYvHkzFy5c0Nft4uJCSEhIlvref/993N3dcXZ2ZurUqUCaDkTv3r31afz9/enatSsAy5Yto169enh7ezN8+HDGjh2bY1+uXbuGl5cXTZo0YcqUKU+Ul3HtIC1U+ooVK4C0mE6ffPIJHh4eeHh4cPXq1Vyvl0EjpXzmC3gTuELaL/7J6cemk2YMAP4AIoHA9Nf29OPNSFvGGpT+/t6z6mrcuLEsag5cipSeX/0ha0/aIb/dfVEmpqQWeRsUJYfg4OBirf/8+fOyXr168v79+1JKKaOioqSUUnbp0kWuWLFCSinlsmXLZPfu3aWUUjZs2FDeunVLSilldHS0lFLKgwcPys6dO0sppZw6dar08vKSSUlJ8v79+7J8+fIyJSVFBgcHyy5dusiUlBQppZTvv/++XLlypZRSyvfee0+eOnUqS9s6d+4s165dK6WUcuHChdLa2jpLfU+zZcsWOWjQICmllMnJybJatWoyISFBjh07Vq5evVp/PCEhIUvejL6npqbK1q1by6CgIKnRaGT16tVlXFyclFLKUaNGSV9fX3n79m1Zs2ZNGRUVJVNSUmSLFi3kmDFjcrzOXbt21fd33rx5OfZlzJgxcvny5VJKKWvWrClnzJghpZRy5cqVOfb5RcnubxAIkHm4n+f1laeorFLKXVLKelLKulLKr9KPfS6l3J7+uZ2U8iUppVv6q1v68WNSylellK7p78sKxKIVEDGJGib6BTFk+SlsLU34dXRzPumovAWFYZObnkPGY5uBAwdy5MgRAJo3b46Pjw9LlizJ8Zd/586dMTc3x97enkqVKhEZGcn+/fs5ffo0TZo0wc3Njf379xMWlqZ6t3Tp0mwfS/311196nYe8PkLq1KkTBw4cIDk5md27d9OqVSssLS3x8vLi66+/ZubMmVy/fh1LS8sseTdu3Mhrr71Go0aNuHDhAsHBwU/oQKSmprJz5066d+/OyZMnad26NeXLl8fU1PQJPYrsOHr0KH379tVfz7ySkadv37789ddfec5naBjsDunC5sClSD7dco4HcSmMaVOXcW0dMTdRRkFh+Mjn1HNYtGgRJ06cYOfOnbi5uWWrw5CdxoGUksGDB2cb9bQgsbCwwNvbmz179rBhwwb9zbVfv354enqyc+dOOnTowNKlS3n99df1+a5du8bs2bM5deoU5cqVw8fHh6SkJCB7HQiZjyCj2V3nzFoRgL7O7PIU1xL8gqDE6Tm8KDEJGv67MYihKwIoa2nG1tHNmdihvjIMihLD8+o5hIaG4unpyfTp07G3t+fmzZvZF5xNPZs2beLevXv6eq5fv55rnqZNm7J582YAfVsgd80GgHfffZfly5dz+PBhOnToAEBYWBh16tRh3LhxdOvWjbNnzz6RJzY2Fmtra+zs7IiMjGT37n9W0GenA+Hh4cGhQ4eIjo4mNTVV386caN68+RPXM4OaNWsSHBxMcnIyMTEx7N+//4l8GzZs0L97eXnlWoch868yDn8ER/LGj4fYGnibD15/me0fNOfVakUfhkOheBEy6zm4urry0UcfATB37lyWL1+Oi4sLvr6+zJkzB4CJEyfy6quv0rBhQ1q1aoWrq2ue6nFycmLGjBm0b98eFxcX3njjDSIiIgAYNmwY2YXW/+mnn/jhhx/w8PAgIiJCr9ng4uKCiYkJrq6uWSakAdq3b8+ff/5Ju3btMDNLE+PasGEDDRs2xM3NjUuXLmVRqXN1daVRo0Y4OzszdOhQmjdvrj+XoQOxe/du/eRx1apV+eyzz/D09KRdu3Y4OTnp25cdc+bMYf78+TRp0oSYmBj98erVq9O7d29cXFzo378/jRo1eiJfcnIynp6ezJkzJ9u+lhRKjJ7Di/AoIYXpvwWz5e/b1K9chtm9XGlYVRkFRf5Qeg45k5CQgKWlJUII1q9fz7p169i2bVtxN0tPhqZEamoqb731FkOHDuWtt94qsPJr1apFQECAfj6osFB6DgXAvuBIPvv1HNHxKYxr68jYNi9jZvKvcpgUiiLj9OnTjB07FiklZcuWzaLtXNxMmzaNP/74g6SkJNq3b0+PHj2Ku0kGS6k1DtHxKXzx2wW2Bt6hgYMty32aKG9BoShkWrZs+Uw96eJk9uzZWY599dVXej3sDHr16sXkyZOfu/zw8PD8Ns3gKJXGYc+Fu0z+9TyPElL4sJ0jo72Vt6BQKLJn8uTJ+TIEpZ1SZRwexqcwbfsFtgfdwcnBllVDPXCqYlvczVIoFIoSR6kxDr+fj+B/W88Tk6jhozfq8b53XUyNlbegUCgU+aHEG4eH8Sl8vu08O85G0LCqLb7vedLAQXkLCoVC8SKUaOOw61wEU7aeJzZJw4T29RjZWnkLCoVCURCUyDtpVFwyY9acYfSaM1Qpa8lvH7Rg7OuOyjAoFHlE6TkUDsOGDSM42CADTz83Jc5z2Hk2ginbzhOXlMrEDq8wslUdTJRRUBQTu3fv5u7duwVaZuXKlenUqVOBlvk0Ss+hcFi6dGlxN6HAKDF31QdxyYxec5oxa89QvZwlO8a1YEybl5VhUPwrUXoOaRSmnoOPjw+bNm3Sf8/oi7+/P97e3vTs2ZP69evTv39/fVA/b2/vbMOKlEgKMv53Qbye1nPQ6XRye+Bt6fbFHun42S654OBVqUnV5hzoXKEoZJSew79Dz2Hw4MHSz89P/z1zX2xtbeXNmzelVquVTZs2lYcPH5ZSStm6detsr0tBYzB6DsXF/cfJvL/6DB+s+5saFazZOa4F73vXVd6C4l+N0nP4h8LUc8gNDw8PqlWrhpGREW5ubqVqZ3QGebrLCiE6CiEuCyGuCiEmZXP+IyFEsBDirBBivxCiZqZzg4UQIemvwXmpT0rJtsDbvPHjIQ5cvsekTvXZPMoLx5fK5L1nCkUpReZDz2HGjBncvHkTNzc3fajvzOSm5xAYGEhgYCCXL19m2rRpBdaPDJ7Wc3j33XeBNOOyfft2LC0t6dChAwcOHHgiX4aew/79+zl79iydO3d+Qs9h48aNHDhwIN96Dpl1G6SUpKSk6M9ld71KG880DkIIY2A+0AlwAvoKIZyeSvY34C6ldAE2AbPS85YHpgKegAcwVQhRLrf6UnWSkb6nGb8+kFoVrNk1rgWjWitvQaHIQOk5pFHYeg61atXi9OnTAGzbtg2NRpNr+tJGXu64HsBVKWWYlDIFWA90z5xASnlQSpmQ/vU4UC39cwdgn5TyoZQyGtgHdMytsiuRjzl05T6fvVmfze834+VKyltQKDKj9BzSKGw9h+HDh3Po0CE8PDw4ceIE1tbWebpupYVn6jkIIXoCHaWUw9K/DwQ8pZTZTvMLIeYBd6WUM4QQEwALKeWM9HNTgEQp5eyn8owARgDYONRpHBgURN2KNi/YNYWicFB6Djnzb9dzKCoMRc8hu4eb2VoUIcQAwB1o/Tx5pZSLgcWQJvajDINCUTJReg6lh7wYh1tA9UzfqwF3nk4khGgHTAZaSymTM+X1fiqvf34aqlAoDJ9/u55DaSIvxuEU4CiEqA3cBt4FnlijJoRoBPxM2uOne5lO7QG+zjQJ3R749IVbrVAUM3ldMaQwfEqansPzrrrKL8+ckJZSpgJjSbvRXwQ2SikvCCGmCyG6pSf7DrAB/IQQgUKI7el5HwJfkmZgTgHT048pFCUWCwsLoqKiiuyfVKHIQEpJVFQUFhYWhV7XMyekixp3d3dZarafK0olGo2GW7du6dfUKxRFiYWFBdWqVcPU1PSJ48UxIa1QKDJhampK7dq1i7sZCkWhonaWKRQKhSILyjgoFAqFIgvKOCgUCoUiCwY3IS2EeAxcLu52FCL2wIPibkQhovpXsinN/SvNfQN4RUpZYPGGDHFC+nJBzrgbGkKIANW/kovqX8mlNPcN0vpXkOWpx0oKhUKhyIIyDgqFQqHIgiEah8XF3YBCRvWvZKP6V3IpzX2DAu6fwU1IKxQKhaL4MUTPQaFQKBTFjDIOCoVCochCkRsHIcQvQoh7QojzmY6VF0LsE0KEpL+XSz8uhBBzhRBXhRBnhRCvFXV7Ff8ghKguhDgohLgohLgghBifflyNXwlBCBEuhDiXHj05IP2YGj8DpaDul0KIwenpQ4QQg/NSd3F4DivIqiM9CdgvpXQE9qd/B+gEOKa/RgALi6iNiuxJBf4rpWwANAXGCCGcUONX0mgjpXTLtOZfjZ/hsoIXvF8KIcoDUwFPwAOYmkljJ0eK3DhIKf8EntZ06A6sTP+8EuiR6fgqmcZxoKwQwqFoWqp4GillhJTyTPrnx6Tpe1RFjV9JR42fgVJA98sOwD4p5UMpZTSwj6wGJwuGMufwkpQyAtJuQECl9ONVgZuZ0t1KP6YoZoQQtYBGwAnU+JUkJLBXCHFaCDEi/Zgav5LF845XvsbREMNnZCY7HUa19raYEULYAJuBD6WUsbnIZarxMzyaSynvCCEqAfuEEJdySavGr2SR03jlaxwNxXOIzHBX098zdKhvAdUzpasG3CnitikyIYQwJc0wrJFSbkk/rMavhCClvJP+fg/4lbRn0Gr8ShbPO175GkdDMQ7bgYwZ9MHAtkzHB6XPwjcFYjLcKUXRI9JchGXARSnlD5lOqfErAQghrIUQZTI+A+2B86jxK2k873jtAdoLIcqlT0S3Tz+WO1LKIn0B64AIQEOaRXsPqEDarHtI+nv59LQCmA+EAucA96Jur3o9MXYtSHNHzwKB6a831fiVjBdQBwhKf10AJqcfV+NnoK+Cul8CQ4Gr6a8healbhc9QKBQKRRYM5bGSQqFQKAwIZRwUCoVCkQVlHBQKhUKRBWUcFAqFQpEFZRwUCoVCkQVlHBQKhUKRBWUcFAqFQpGF/weBK2vnS/NxUwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "values_10 = [k_10_100_m, k_10_200_m, k_10_500_m, k_10_1000_m]\n",
    "create_df_line(values_10, 10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "***\n",
    "## Top 4 graphs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:25:10.015419Z",
     "start_time": "2019-08-13T23:25:10.000956Z"
    }
   },
   "outputs": [],
   "source": [
    "def tanimoto_euclidean_2(moldb, k, radius):\n",
    "    lst1, lst2 = [], []\n",
    "    query_size = 20\n",
    "    \n",
    "    for i in range(query_size):\n",
    "        gt_rankings = tanimoto_similarity(moldb, i)\n",
    "        #dup_rankings = mol2vec_euclidean_similarity(moldb, i, radius, mol2vec_duplicates)\n",
    "        uni_rankings = mol2vec_euclidean_similarity(moldb, i, radius, mol2vec_unique)\n",
    "        #avg_dup_rankings = mol2vec_euclidean_similarity(moldb, i, radius, mol2vec_average_dup)\n",
    "        avg_uni_rankings = mol2vec_euclidean_similarity(moldb, i, radius, mol2vec_average_uni)\n",
    "\n",
    "        lst1.append(compare_rankings_recall(gt_rankings, uni_rankings, k))\n",
    "        lst2.append(compare_rankings_recall(gt_rankings, avg_uni_rankings, k))\n",
    "    \n",
    "    return [lst1, lst2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:25:10.568347Z",
     "start_time": "2019-08-13T23:25:10.562063Z"
    }
   },
   "outputs": [],
   "source": [
    "def tanimoto_cosine_2(moldb, k, radius):\n",
    "    lst1, lst2 = [], []\n",
    "    query_size = 20\n",
    "\n",
    "    for i in range(query_size):\n",
    "        gt_rankings = tanimoto_similarity(moldb, i)\n",
    "        #dup_rankings = mol2vec_cosine_similarity(moldb, i, radius, mol2vec_duplicates)\n",
    "        uni_rankings = mol2vec_cosine_similarity(moldb, i, radius, mol2vec_unique)\n",
    "        #avg_dup_rankings = mol2vec_cosine_similarity(moldb, i, radius, mol2vec_average_dup)\n",
    "        avg_uni_rankings = mol2vec_cosine_similarity(moldb, i, radius, mol2vec_average_uni)\n",
    "\n",
    "        lst1.append(compare_rankings_recall(gt_rankings, uni_rankings, k))\n",
    "        lst2.append(compare_rankings_recall(gt_rankings, avg_uni_rankings, k))\n",
    "    \n",
    "    return [lst1, lst2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:28:49.057639Z",
     "start_time": "2019-08-13T23:25:11.381578Z"
    }
   },
   "outputs": [],
   "source": [
    "# k=10\n",
    "# put 8 functions together for all sample sizes\n",
    "k_10_100 = tanimoto_euclidean_2(c_small_moldb100, 10, 1)+tanimoto_cosine_2(c_small_moldb100, 10, 1)\n",
    "k_10_200 = tanimoto_euclidean_2(c_small_moldb200, 10, 1)+tanimoto_cosine_2(c_small_moldb200, 10, 1)\n",
    "k_10_500 = tanimoto_euclidean_2(c_small_moldb500, 10, 1)+tanimoto_cosine_2(c_small_moldb500, 10, 1)\n",
    "k_10_1000 = tanimoto_euclidean_2(c_small_moldb1000, 10, 1)+tanimoto_cosine_2(c_small_moldb1000, 10, 1)\n",
    "k_10_2000 = tanimoto_euclidean_2(c_small_moldb2000, 10, 1)+tanimoto_cosine_2(c_small_moldb2000, 10, 1)\n",
    "k_10_5000 = tanimoto_euclidean_2(c_small_moldb5000, 10, 1)+tanimoto_cosine_2(c_small_moldb5000, 10, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:43:49.341943Z",
     "start_time": "2019-08-13T23:43:49.327942Z"
    }
   },
   "outputs": [],
   "source": [
    "# takes in a list of lists, returns a list of median of all lists\n",
    "def med(lst):\n",
    "    med_lst = []\n",
    "    \n",
    "    for i in lst:\n",
    "        med = np.median(i)\n",
    "        med_lst.append(med)\n",
    "    \n",
    "    return med_lst"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:43:50.317297Z",
     "start_time": "2019-08-13T23:43:50.302413Z"
    }
   },
   "outputs": [],
   "source": [
    "# find median\n",
    "k_10_100_m = med(k_10_100)\n",
    "k_10_200_m = med(k_10_200)\n",
    "k_10_500_m = med(k_10_500)\n",
    "k_10_1000_m = med(k_10_1000)\n",
    "k_10_2000_m = med(k_10_2000)\n",
    "k_10_5000_m = med(k_10_5000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:43:55.408469Z",
     "start_time": "2019-08-13T23:43:55.402267Z"
    }
   },
   "outputs": [],
   "source": [
    "# put into df\n",
    "values = [k_10_100_m, k_10_200_m, k_10_500_m, k_10_1000_m, k_10_2000_m, k_10_5000_m]\n",
    "cols = ['euclidean: gt vs uni', 'euclidean: gt vs avg_uni', \n",
    "       'cosine: gt vs uni', 'cosine: gt vs avg_uni']\n",
    "df = pd.DataFrame(values, index=['100', '200', '500', '1000', '2000', '5000'], columns=cols)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-08-13T23:43:57.597878Z",
     "start_time": "2019-08-13T23:43:57.211953Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7ff9f20e4358>"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "ls = [':', '-', '--', '-.']\n",
    "fig, ax = plt.subplots()\n",
    "df.plot(y=0, kind='line', ax=ax, linestyle=ls[1], linewidth=3, alpha=0.5)\n",
    "df.plot(y=1, kind='line', ax=ax, linestyle=ls[0], linewidth=3, alpha=0.6)\n",
    "df.plot(y=2, kind='line', ax=ax, linestyle=ls[2], linewidth=3, alpha=0.5)\n",
    "df.plot(y=3, kind='line', ax=ax, linestyle=ls[0], linewidth=3, alpha=0.5)\n",
    "plt.title('Top 4 distances: k=10')\n",
    "plt.legend(loc='lower right')"
   ]
  },
  {
   "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}