Step72.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 281,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The autoreload extension is already loaded. To reload it, use:\n",
      "  %reload_ext autoreload\n"
     ]
    }
   ],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "import matplotlib.pyplot as plt\n",
    "from Environment.E_commerce import *\n",
    "from Environment.Product import *\n",
    "from Algorithms.Conv_rates_Learner import *\n",
    "from Algorithms.TS_Learner_poisson import *\n",
    "from Algorithms.CG_Learner_Environment import *\n",
    "from Algorithms.UCB_Learner import *\n",
    "from Algorithms.Greedy_Learner import *\n",
    "from Algorithms.CG_UCB_Learner import *\n",
    "import copy"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Context generation environment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 282,
   "metadata": {},
   "outputs": [],
   "source": [
    "P1 = Product(0,[19.99, 21.99, 23.99, 26.99],np.multiply([0.2, 0.2, 0.2, 0.2],[19.99, 19.99, 19.99, 19.99])+[0, 21.99-19.99, 23.99-19.99, 26.99-19.99])\n",
    "# t-shirt\n",
    "P2 = Product(1,[29.99, 32.59, 35.99, 39.99],np.multiply([0.2, 0.2, 0.2, 0.2],[29.99, 29.99, 29.99, 29.99])+[29.99, 32.59, 35.99, 39.99]-29.99)\n",
    "# shirt\n",
    "P3 = Product(2,[25.99,28.99, 31.99, 34.99],np.multiply([0.3, 0.3, 0.3, 0.3],[25.99,25.99, 25.99, 25.99])+[25.99,28.99, 31.99, 34.99]-25.99)\n",
    "# shorts\n",
    "P4 = Product(3,[38.99,43.99,46.99, 49.99],np.multiply([0.4, 0.4, 0.4, 0.4],[38.99,38.99,38.99, 38.99])+[38.99,43.99,46.99, 49.99]-38.99)\n",
    "# pants\n",
    "P5 = Product(4,[79.99,89.99, 99.99, 109.99],np.multiply([0.5, 0.5, 0.5, 0.5],[79.99,79.99, 79.99, 79.99])+[79.99,89.99, 99.99, 109.99]-79.99)\n",
    "# shoes\n",
    "products = [P1,P2,P3,P4,P5]\n",
    "margins_matrix = np.zeros((5,4))\n",
    "expected_units_sold_per_product = np.array([2,1,3,3,1]) #theese are the lambdas defined in the visit function\n",
    "for i in range(5):\n",
    "    for j in range(4):\n",
    "        margins_matrix[i,j] = products[i].margins_list[j]\n",
    "E = CG_E_commerce()\n",
    "E.set_lambda(0.5)\n",
    "for f1 in range(2):\n",
    "    for f2 in range(2):\n",
    "        E.set_products(f1,f2,products)\n",
    "E.graph = np.array([[0. , 0.5, 0. , 1. , 0. ],[1. , 0. , 0. , 0.5, 0. ],[0. , 1. , 0. , 0.5, 0. ],[0. , 0.5, 1. , 0. , 0. ],[1. , 0. , 0. , 0.5, 0. ]])\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Context Generation Algorithm (TS and Random Algorithm)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 283,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "period 0\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\foxfo\\AppData\\Roaming\\Python\\Python39\\site-packages\\numpy\\core\\fromnumeric.py:86: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.\n",
      "  return ufunc.reduce(obj, axis, dtype, out, **passkwargs)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mu_c0_cr= [[1021.4347706278705, 0], [0, 0]]\n",
      "mu_c0_ts= [[1281.734282657559, 0], [0, 0]]\n",
      "period 1\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 1038.4734929776614\n",
      "mu_c0_cr= [[1021.4347706278705, 0], [1038.4734929776614, 0]]\n",
      "--- SPLIT F1 MAKES SENSE ---\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 2132.6857496701027\n",
      "mu_c0_ts= [[1281.734282657559, 0], [2132.6857496701027, 0]]\n",
      "--- SPLIT F1 MAKES SENSE ---\n",
      "period 2\n",
      "mu_c0_cr= [[1021.4347706278705, 0], [1038.4734929776614, 991.8532802382032]]\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 991.8532802382032\n",
      "--- SPLIT F2 DOESN'T MAKE SENSE (we consider only f1)---\n",
      "mu_c0_ts= [[1281.734282657559, 0], [2132.6857496701027, 2247.681508868027]]\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 2247.681508868027\n",
      "--- SPLIT F2 MAKES SENSE (together with f1)---\n",
      "period 3\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 1222.1105945963054\n",
      "mu_c0_cr= [[1021.4347706278705, 0], [1222.1105945963054, 991.8532802382032]]\n",
      "--- SPLIT F1 MAKES SENSE ---\n",
      "mu_c0_ts= [[1281.734282657559, 0], [2132.6857496701027, 4866.683428498838]]\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 4866.683428498838\n",
      "--- SPLIT F2 MAKES SENSE (together with f1)---\n",
      "period 4\n",
      "mu_c0_cr= [[1021.4347706278705, 0], [1222.1105945963054, 1359.3410038806167]]\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 1359.3410038806167\n",
      "--- SPLIT F2 MAKES SENSE (together with f1)---\n",
      "mu_c0_ts= [[1281.734282657559, 0], [2132.6857496701027, 7448.713860669336]]\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 7448.713860669336\n",
      "--- SPLIT F2 MAKES SENSE (together with f1)---\n",
      "period 5\n",
      "mu_c0_cr= [[1021.4347706278705, 0], [1222.1105945963054, 2182.774070416205]]\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 2182.774070416205\n",
      "--- SPLIT F2 MAKES SENSE (together with f1)---\n",
      "mu_c0_ts= [[1281.734282657559, 0], [2132.6857496701027, 9894.304772831814]]\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 9894.304772831814\n",
      "--- SPLIT F2 MAKES SENSE (together with f1)---\n",
      "period 6\n",
      "mu_c0_cr= [[1021.4347706278705, 0], [1222.1105945963054, 3270.014683812632]]\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 3270.014683812632\n",
      "--- SPLIT F2 MAKES SENSE (together with f1)---\n",
      "mu_c0_ts= [[1281.734282657559, 0], [2132.6857496701027, 12356.485588876267]]\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 12356.485588876267\n",
      "--- SPLIT F2 MAKES SENSE (together with f1)---\n",
      "period 7\n",
      "mu_c0_cr= [[1021.4347706278705, 0], [1222.1105945963054, 4477.084283553343]]\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 4477.084283553343\n",
      "--- SPLIT F2 MAKES SENSE (together with f1)---\n",
      "mu_c0_ts= [[1281.734282657559, 0], [2132.6857496701027, 14735.566377202154]]\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 14735.566377202154\n",
      "--- SPLIT F2 MAKES SENSE (together with f1)---\n",
      "period 8\n",
      "mu_c0_cr= [[1021.4347706278705, 0], [1222.1105945963054, 5562.024671344838]]\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 5562.024671344838\n",
      "--- SPLIT F2 MAKES SENSE (together with f1)---\n",
      "mu_c0_ts= [[1281.734282657559, 0], [2132.6857496701027, 17361.568953887603]]\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 17361.568953887603\n",
      "--- SPLIT F2 MAKES SENSE (together with f1)---\n",
      "period 9\n",
      "mu_c0_cr= [[1021.4347706278705, 0], [1222.1105945963054, 6899.5932321788805]]\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 6899.5932321788805\n",
      "--- SPLIT F2 MAKES SENSE (together with f1)---\n",
      "mu_c0_ts= [[1281.734282657559, 0], [2132.6857496701027, 19751.843617207618]]\n",
      "p_c1*mu_c1 + p_c2*mu_c2= 19751.843617207618\n",
      "--- SPLIT F2 MAKES SENSE (together with f1)---\n"
     ]
    }
   ],
   "source": [
    "#Context generation for TS and random (reset every 2 weeks)\n",
    "\n",
    "np.random.seed(1000)\n",
    "n_arms = 4\n",
    "num_users = 100\n",
    "n_days = 14*10\n",
    "fixed_alpha = 0\n",
    "fixed_weights = 1\n",
    "fixed_units = 0\n",
    "num_experiments = 1\n",
    "opt_vector = [[np.zeros(num_experiments) for _ in range(2)] for _ in range(2)]\n",
    "conv_rates_per_experiment = [[[] for _ in range(2)] for _ in range(2)]\n",
    "mean_units_sold_per_product_per_eperiment = [[[] for _ in range(2)] for _ in range(2)]\n",
    "cr_learner_expected_rewards_per_experiment = [[[] for _ in range(2)] for _ in range(2)]\n",
    "collected_reward_crs = np.array([])\n",
    "collected_reward_tss = np.array([])\n",
    "\n",
    "mu_c0_cr = [[0 for _ in range(2)] for _ in range(2)]\n",
    "mu_c0_ts = [[0 for _ in range(2)] for _ in range(2)]\n",
    "for e in range(num_experiments):\n",
    "    env= CG_Environment(n_arms, E, margins_matrix, num_users, fixed_alpha, fixed_weights,fixed_units)\n",
    "\n",
    "    ## LEARN \\mu0\n",
    "    cr_learner = Conv_rates(n_arms=n_arms)\n",
    "    ts_learner = TS_poisson(n_arms=n_arms)\n",
    "\n",
    "    pulled_arm_cr = [[np.zeros(5) for _ in range(2)] for _ in range(2)]\n",
    "    pulled_arm_ts = [[np.zeros(5) for _ in range(2)] for _ in range(2)]\n",
    "\n",
    "    split1_cr = 0\n",
    "    split2_cr = 0\n",
    "\n",
    "    split1_ts = 0\n",
    "    split2_ts = 0\n",
    "\n",
    "    for period in range(int(n_days/14)):\n",
    "        print(\"period\",period)\n",
    "        for d in range(14):\n",
    "            if split1_cr==0 and split2_cr==0:\n",
    "                for f1 in range(2):\n",
    "                    for f2 in range(2):\n",
    "                        pulled_arm_cr[f1][f2] = cr_learner.pull_arm()\n",
    "\n",
    "                reward_cr = env.round(pulled_arm_cr)  #reward_cr[f1][f2][i]\n",
    "                cr_learner.update(pulled_arm_cr[0][0] , reward_cr[0][0], env.clicks_current_day[0][0], env.purchases_current_day[0][0],env.daily_units[0][0])\n",
    "\n",
    "            # FIRST DECISION\n",
    "            if split1_cr==1 and split2_cr==0:\n",
    "                pulled_arm_cr = [[np.zeros(5) for _ in range(2)] for _ in range(2)]\n",
    "                for f1 in range(2):\n",
    "                    for f2 in range(2):\n",
    "                        pulled_arm_cr[f1][f2] = cr_learners[f1].pull_arm()\n",
    "\n",
    "                reward_cr = env.round(pulled_arm_cr)\n",
    "\n",
    "                for f1 in range(2):\n",
    "                    cr_learners[f1].update(pulled_arm_cr[f1][0], (reward_cr[f1][0]+reward_cr[f1][1]), (env.clicks_current_day[f1][0]+env.clicks_current_day[f1][1]), (env.purchases_current_day[f1][0]+env.purchases_current_day[f1][1]),(env.daily_units[f1][0]+env.daily_units[f1][1]))\n",
    "\n",
    "            elif split1_cr==1 and split2_cr==1:  # we add the second split\n",
    "                pulled_arm_cr = [[np.zeros(5) for _ in range(2)] for _ in range(2)]\n",
    "                for f1 in range(2):\n",
    "                    for f2 in range(2):\n",
    "                        pulled_arm_cr[f1][f2] = cr_learners[f1][f2].pull_arm()\n",
    "\n",
    "                reward_cr = env.round(pulled_arm_cr)\n",
    "\n",
    "                for f1 in range(2):\n",
    "                    for f2 in range(2):\n",
    "                        cr_learners[f1][f2].update(pulled_arm_cr[f1][f2], reward_cr[f1][f2], env.clicks_current_day[f1][f2], env.purchases_current_day[f1][f2],env.daily_units[f1][f2])\n",
    "\n",
    "            elif split1_cr == 0 and split2_cr ==1:\n",
    "                pulled_arm_cr = [[np.zeros(5) for _ in range(2)] for _ in range(2)]\n",
    "                for f1 in range(2):\n",
    "                    for f2 in range(2):\n",
    "                        pulled_arm_cr[f1][f2] = cr_learners[f2].pull_arm()\n",
    "\n",
    "                reward_cr = env.round(pulled_arm_cr)\n",
    "\n",
    "                for f2 in range(2):\n",
    "                    cr_learners[f2].update(pulled_arm_cr[0][f2], (reward_cr[0][f2]+reward_cr[1][f2]), (env.clicks_current_day[0][f2]+env.clicks_current_day[1][f2]), (env.purchases_current_day[0][f2]+env.purchases_current_day[1][f2]),(env.daily_units[0][f2]+env.daily_units[1][f2]))\n",
    "\n",
    "        # --------------------------------------------------\n",
    "        # -------------------THOMSON SAMPLING ---------------\n",
    "        # --------------------------------------------------\n",
    "        #ciao\n",
    "\n",
    "            if split1_ts==0 and split2_ts==0:\n",
    "                    for f1 in range(2):\n",
    "                        for f2 in range(2):\n",
    "                            pulled_arm_ts[f1][f2] = ts_learner.pull_arm(margins_matrix)\n",
    "\n",
    "                    reward_ts = env.round(pulled_arm_ts)  #reward_ts[f1][f2][i]\n",
    "                    ts_learner.update(pulled_arm_ts[0][0] , reward_ts[0][0], env.clicks_current_day[0][0], env.purchases_current_day[0][0],env.daily_units[0][0])\n",
    "\n",
    "                # FIRST DECISION\n",
    "            if split1_ts==1 and split2_ts==0:\n",
    "                pulled_arm_ts = [[np.zeros(5) for _ in range(2)] for _ in range(2)]\n",
    "                for f1 in range(2):\n",
    "                    for f2 in range(2):\n",
    "                        pulled_arm_ts[f1][f2] = ts_learners[f1].pull_arm(margins_matrix)\n",
    "\n",
    "                reward_ts = env.round(pulled_arm_ts)\n",
    "\n",
    "                for f1 in range(2):\n",
    "                    ts_learners[f1].update(pulled_arm_ts[f1][0], (reward_ts[f1][0]+reward_ts[f1][1]), (env.clicks_current_day[f1][0]+env.clicks_current_day[f1][1]), (env.purchases_current_day[f1][0]+env.purchases_current_day[f1][1]),(env.daily_units[f1][0]+env.daily_units[f1][1]))\n",
    "\n",
    "            elif split1_ts==1 and split2_ts==1:  # we add the second split\n",
    "                pulled_arm_ts = [[np.zeros(5) for _ in range(2)] for _ in range(2)]\n",
    "                for f1 in range(2):\n",
    "                    for f2 in range(2):\n",
    "                        pulled_arm_ts[f1][f2] = ts_learners[f1][f2].pull_arm(margins_matrix)\n",
    "\n",
    "                reward_ts = env.round(pulled_arm_ts)\n",
    "\n",
    "                for f1 in range(2):\n",
    "                    for f2 in range(2):\n",
    "                        ts_learners[f1][f2].update(pulled_arm_ts[f1][f2], reward_ts[f1][f2], env.clicks_current_day[f1][f2], env.purchases_current_day[f1][f2],env.daily_units[f1][f2])\n",
    "\n",
    "            elif split1_ts == 0 and split2_ts ==1:\n",
    "                pulled_arm_ts = [[np.zeros(5) for _ in range(2)] for _ in range(2)]\n",
    "                for f1 in range(2):\n",
    "                    for f2 in range(2):\n",
    "                        pulled_arm_ts[f1][f2] = ts_learners[f2].pull_arm(margins_matrix)\n",
    "\n",
    "                reward_ts = env.round(pulled_arm_ts)\n",
    "\n",
    "                for f2 in range(2):\n",
    "                    ts_learners[f2].update(pulled_arm_ts[0][f2], (reward_ts[0][f2]+reward_ts[1][f2]), (env.clicks_current_day[0][f2]+env.clicks_current_day[1][f2]), (env.purchases_current_day[0][f2]+env.purchases_current_day[1][f2]),(env.daily_units[0][f2]+env.daily_units[1][f2]))\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "        # DECISION\n",
    "        #compute sample probabilities and make decision\n",
    "        if split1_cr==0 and split2_cr==0:\n",
    "            mu_c0_cr[split1_cr][split2_cr] = np.sum(np.sum(cr_learner.rewards_per_arm))\n",
    "            print('mu_c0_cr=',mu_c0_cr)\n",
    "             # split 1\n",
    "            collected_reward_crs=np.concatenate((collected_reward_crs, cr_learner.collected_rewards)) #axis=None\n",
    "            #cr_learners = [cr_learner,cr_learner]\n",
    "            cr_learners = [Conv_rates(n_arms=n_arms),Conv_rates(n_arms=n_arms)]\n",
    "            split1_cr = 1\n",
    "            split2_cr = 0\n",
    "\n",
    "        elif split1_cr==1 and split2_cr==0:\n",
    "            P = env.E.users_by_class/np.sum(env.E.users_by_class)\n",
    "            p_c1 = P[0][0] + P[0][1]\n",
    "            p_c2 = P[1][0] + P[1][1]\n",
    "            mu_c1 = np.sum(np.sum(cr_learners[0].rewards_per_arm))\n",
    "            mu_c2 = np.sum(np.sum(cr_learners[1].rewards_per_arm))\n",
    "\n",
    "            # we consider the second split\n",
    "            mu_c0_cr[split1_cr][split2_cr] = p_c1*mu_c1 + p_c2*mu_c2\n",
    "            split2_cr = 1  # in ogni caso consideremo la seconds feature\n",
    "            print('p_c1*mu_c1 + p_c2*mu_c2=',p_c1*mu_c1 + p_c2*mu_c2)\n",
    "            print('mu_c0_cr=',mu_c0_cr)\n",
    "            collected_reward_crs=np.concatenate((collected_reward_crs,\n",
    "                                                  cr_learners[0].collected_rewards + cr_learners[1].collected_rewards)) #axis=None\n",
    "            if p_c1*mu_c1 + p_c2*mu_c2 > mu_c0_cr[0][0]:\n",
    "                cr_learners = [[Conv_rates(n_arms=n_arms),Conv_rates(n_arms=n_arms)] for _ in range(2)]\n",
    "                split1_cr = 1\n",
    "                print(\"--- SPLIT F1 MAKES SENSE ---\")\n",
    "            else:\n",
    "                cr_learners = [Conv_rates(n_arms=n_arms),Conv_rates(n_arms=n_arms)]\n",
    "                split1_cr = 0\n",
    "                print(\"--- SPLIT F1 DOESN'T MAKE SENSE ---\")\n",
    "\n",
    "        elif split1_cr==1 and split2_cr==1:\n",
    "            P = env.E.users_by_class/np.sum(env.E.users_by_class)\n",
    "            p_c1 = P[0][0] + P[1][0]  #p(seconda feature = 0)\n",
    "            p_c2 = P[0][1] + P[1][1]  #p(seconda feature = 1)\n",
    "            mu_c1 = np.sum(np.sum(cr_learners[0][0].rewards_per_arm))+np.sum(np.sum(cr_learners[1][0].rewards_per_arm))\n",
    "            mu_c2 = np.sum(np.sum(cr_learners[0][1].rewards_per_arm))+np.sum(np.sum(cr_learners[1][1].rewards_per_arm))\n",
    "\n",
    "            mu_c0_cr[split1_cr][split2_cr] = p_c1*mu_c1 + p_c2*mu_c2\n",
    "            collected_reward_crs=np.concatenate((collected_reward_crs,\n",
    "                                                  cr_learners[0][0].collected_rewards +\n",
    "                                              cr_learners[0][1].collected_rewards +\n",
    "                                              cr_learners[1][0].collected_rewards +\n",
    "                                              cr_learners[1][1].collected_rewards )\n",
    "                                              ) #axis=None\n",
    "\n",
    "            print('mu_c0_cr=',mu_c0_cr)\n",
    "            print('p_c1*mu_c1 + p_c2*mu_c2=',p_c1*mu_c1 + p_c2*mu_c2)\n",
    "\n",
    "            if p_c1*mu_c1 + p_c2*mu_c2 > mu_c0_cr[1][0]:\n",
    "                mu_c0_cr[split1_cr][split2_cr] = p_c1*mu_c1 + p_c2*mu_c2\n",
    "                print(\"--- SPLIT F2 MAKES SENSE (together with f1)---\")\n",
    "            else:\n",
    "                cr_learners = [Conv_rates(n_arms=n_arms),Conv_rates(n_arms=n_arms)]\n",
    "                split2_cr = 0\n",
    "                print(\"--- SPLIT F2 DOESN'T MAKE SENSE (we consider only f1)---\")\n",
    "\n",
    "        elif split1_cr == 0 and split2_cr ==1:\n",
    "            P = env.E.users_by_class/np.sum(env.E.users_by_class)\n",
    "            p_c1 = P[0][0] + P[1][0]  #p(seconda feature = 0)\n",
    "            p_c2 = P[0][1] + P[1][1]  #p(seconda feature = 1)\n",
    "            mu_c1 = np.sum(np.sum(cr_learners[0].rewards_per_arm))\n",
    "            mu_c2 = np.sum(np.sum(cr_learners[1].rewards_per_arm))\n",
    "\n",
    "            mu_c0_cr[split1_cr][split2_cr] = p_c1*mu_c1 + p_c2*mu_c2\n",
    "\n",
    "            # we consider the second split\n",
    "            print('p_c1*mu_c1 + p_c2*mu_c2=',p_c1*mu_c1 + p_c2*mu_c2)\n",
    "            print('mu_c0_cr=',mu_c0_cr)\n",
    "            collected_reward_crs=np.concatenate((collected_reward_crs,\n",
    "                                                  cr_learners[0].collected_rewards + cr_learners[1].collected_rewards)) #axis=None\n",
    "\n",
    "            if p_c1*mu_c1 + p_c2*mu_c2 > mu_c0_cr[0][0]:\n",
    "                mu_c0_cr[split1_cr][split2_cr] = p_c1*mu_c1 + p_c2*mu_c2\n",
    "                print(\"--- SPLIT F2 MAKES SENSE (without f1)---\")\n",
    "                #cr_learners = [Conv_rates(n_arms=n_arms),Conv_rates(n_arms=n_arms)]\n",
    "            else:\n",
    "                cr_learners = [Conv_rates(n_arms=n_arms)]\n",
    "                split2_cr = 0\n",
    "                print(\"--- SPLIT F2 DOESN'T MAKE SENSE (without f1) ---\")\n",
    "\n",
    "        \n",
    "            # DECISION\n",
    "        #compute sample probabilities and make decision\n",
    "        if split1_ts==0 and split2_ts==0:\n",
    "            mu_c0_ts[split1_ts][split2_ts] = np.sum(np.sum(ts_learner.rewards_per_arm))\n",
    "            print('mu_c0_ts=',mu_c0_ts)\n",
    "             # split 1\n",
    "            collected_reward_tss=np.concatenate((collected_reward_tss, ts_learner.collected_rewards)) #axis=None\n",
    "            #ts_learners = [ts_learner,ts_learner]\n",
    "            ts_learners = [TS_poisson(n_arms=n_arms),TS_poisson(n_arms=n_arms)]\n",
    "            split1_ts = 1\n",
    "            split2_ts = 0\n",
    "\n",
    "        elif split1_ts==1 and split2_ts==0:\n",
    "            P = env.E.users_by_class/np.sum(env.E.users_by_class)\n",
    "            p_c1 = P[0][0] + P[0][1]\n",
    "            p_c2 = P[1][0] + P[1][1]\n",
    "            mu_c1 = np.sum(np.sum(ts_learners[0].rewards_per_arm))\n",
    "            mu_c2 = np.sum(np.sum(ts_learners[1].rewards_per_arm))\n",
    "\n",
    "            # we consider the second split\n",
    "            mu_c0_ts[split1_ts][split2_ts] = p_c1*mu_c1 + p_c2*mu_c2\n",
    "            split2_ts = 1  # in ogni caso consideremo la seconds feature\n",
    "            print('p_c1*mu_c1 + p_c2*mu_c2=',p_c1*mu_c1 + p_c2*mu_c2)\n",
    "            print('mu_c0_ts=',mu_c0_ts)\n",
    "            collected_reward_tss=np.concatenate((collected_reward_tss,\n",
    "                                                  ts_learners[0].collected_rewards + ts_learners[1].collected_rewards)) #axis=None\n",
    "            if p_c1*mu_c1 + p_c2*mu_c2 > mu_c0_ts[0][0]:\n",
    "                ts_learners = [[TS_poisson(n_arms=n_arms),TS_poisson(n_arms=n_arms)] for _ in range(2)]\n",
    "                split1_ts = 1\n",
    "                print(\"--- SPLIT F1 MAKES SENSE ---\")\n",
    "            else:\n",
    "                ts_learners = [TS_poisson(n_arms=n_arms),TS_poisson(n_arms=n_arms)]\n",
    "                split1_ts = 0\n",
    "                print(\"--- SPLIT F1 DOESN'T MAKE SENSE ---\")\n",
    "\n",
    "        elif split1_ts==1 and split2_ts==1:\n",
    "            P = env.E.users_by_class/np.sum(env.E.users_by_class)\n",
    "            p_c1 = P[0][0] + P[1][0]  #p(seconda feature = 0)\n",
    "            p_c2 = P[0][1] + P[1][1]  #p(seconda feature = 1)\n",
    "            mu_c1 = np.sum(np.sum(ts_learners[0][0].rewards_per_arm))+np.sum(np.sum(ts_learners[1][0].rewards_per_arm))\n",
    "            mu_c2 = np.sum(np.sum(ts_learners[0][1].rewards_per_arm))+np.sum(np.sum(ts_learners[1][1].rewards_per_arm))\n",
    "\n",
    "            mu_c0_ts[split1_ts][split2_ts] = p_c1*mu_c1 + p_c2*mu_c2\n",
    "            collected_reward_tss=np.concatenate((collected_reward_tss,\n",
    "                                                  ts_learners[0][0].collected_rewards +\n",
    "                                              ts_learners[0][1].collected_rewards +\n",
    "                                              ts_learners[1][0].collected_rewards +\n",
    "                                              ts_learners[1][1].collected_rewards )\n",
    "                                              ) #axis=None\n",
    "\n",
    "\n",
    "            print('mu_c0_ts=',mu_c0_ts)\n",
    "            print('p_c1*mu_c1 + p_c2*mu_c2=',p_c1*mu_c1 + p_c2*mu_c2)\n",
    "\n",
    "            if p_c1*mu_c1 + p_c2*mu_c2 > mu_c0_ts[1][0]:\n",
    "                mu_c0_ts[split1_ts][split2_ts] = p_c1*mu_c1 + p_c2*mu_c2\n",
    "                print(\"--- SPLIT F2 MAKES SENSE (together with f1)---\")\n",
    "            else:\n",
    "                ts_learners = [TS_poisson(n_arms=n_arms),TS_poisson(n_arms=n_arms)]\n",
    "                split2_ts = 0\n",
    "                print(\"--- SPLIT F2 DOESN'T MAKE SENSE (we consider only f1)---\")\n",
    "\n",
    "        elif split1_ts == 0 and split2_ts ==1:\n",
    "            P = env.E.users_by_class/np.sum(env.E.users_by_class)\n",
    "            p_c1 = P[0][0] + P[1][0]  #p(seconda feature = 0)\n",
    "            p_c2 = P[0][1] + P[1][1]  #p(seconda feature = 1)\n",
    "            mu_c1 = np.sum(np.sum(ts_learners[0].rewards_per_arm))\n",
    "            mu_c2 = np.sum(np.sum(ts_learners[1].rewards_per_arm))\n",
    "\n",
    "            mu_c0_ts[split1_ts][split2_ts] = p_c1*mu_c1 + p_c2*mu_c2\n",
    "\n",
    "            # we consider the second split\n",
    "            print('p_c1*mu_c1 + p_c2*mu_c2=',p_c1*mu_c1 + p_c2*mu_c2)\n",
    "            print('mu_c0_ts=',mu_c0_ts)\n",
    "            collected_reward_tss=np.concatenate((collected_reward_tss,\n",
    "                                                  ts_learners[0].collected_rewards + ts_learners[1].collected_rewards)) #axis=None\n",
    "\n",
    "            if p_c1*mu_c1 + p_c2*mu_c2 > mu_c0_ts[0][0]:\n",
    "                mu_c0_ts[split1_ts][split2_ts] = p_c1*mu_c1 + p_c2*mu_c2\n",
    "                print(\"--- SPLIT F2 MAKES SENSE (without f1)---\")\n",
    "                #ts_learners = [Conv_rates(n_arms=n_arms),Conv_rates(n_arms=n_arms)]\n",
    "            else:\n",
    "                ts_learners = [TS_poisson(n_arms=n_arms)]\n",
    "                split2_ts = 0\n",
    "                print(\"--- SPLIT F2 DOESN'T MAKE SENSE (without f1) ---\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 284,
   "metadata": {},
   "outputs": [],
   "source": [
    "#TS split 2 makes sense\n",
    "#Random algorithm split 1 makes sense"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Basic Environment to estimate Rewards and Regret"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 285,
   "metadata": {},
   "outputs": [],
   "source": [
    "P1 = Product(0,[19.99, 21.99, 23.99, 26.99],np.multiply([0.2, 0.2, 0.2, 0.2],[19.99, 19.99, 19.99, 19.99])+[0, 21.99-19.99, 23.99-19.99, 26.99-19.99])\n",
    "# t-shirt\n",
    "P2 = Product(1,[29.99, 32.59, 35.99, 39.99],np.multiply([0.2, 0.2, 0.2, 0.2],[29.99, 29.99, 29.99, 29.99])+[29.99, 32.59, 35.99, 39.99]-29.99)\n",
    "# shirt\n",
    "P3 = Product(2,[25.99,28.99, 31.99, 34.99],np.multiply([0.3, 0.3, 0.3, 0.3],[25.99,25.99, 25.99, 25.99])+[25.99,28.99, 31.99, 34.99]-25.99)\n",
    "# shorts\n",
    "P4 = Product(3,[38.99,43.99,46.99, 49.99],np.multiply([0.4, 0.4, 0.4, 0.4],[38.99,38.99,38.99, 38.99])+[38.99,43.99,46.99, 49.99]-38.99)\n",
    "# pants\n",
    "P5 = Product(4,[79.99,89.99, 99.99, 109.99],np.multiply([0.5, 0.5, 0.5, 0.5],[79.99,79.99, 79.99, 79.99])+[79.99,89.99, 99.99, 109.99]-79.99)\n",
    "# shoes\n",
    "products = [P1,P2,P3,P4,P5]\n",
    "margins_matrix = np.zeros((5,4))\n",
    "expected_units_sold_per_product = np.array([2,1,3,3,1]) #theese are the lambdas defined in the visit function\n",
    "for i in range(5):\n",
    "    for j in range(4):\n",
    "        margins_matrix[i,j] = products[i].margins_list[j]\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Estimate Conversion Rates considering simultaneously (with one conv_rates learner) the three classes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 286,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_______________________________________________\n",
      "Conversion rates\n",
      "[[0.81677088 0.38125224 0.0828489  0.00165635]\n",
      " [0.86552637 0.47886545 0.07172189 0.00081677]\n",
      " [0.93928716 0.50829959 0.0863802  0.00249167]\n",
      " [0.9563267  0.52693158 0.17428861 0.02206372]\n",
      " [0.99761232 0.73719404 0.36974465 0.02542254]]\n",
      "_______________________________________________\n",
      "Expected rewards per arm\n",
      "[[ 6.5176886   4.59590979  1.52390724  0.01317368]\n",
      " [ 5.0949416   4.1286748   0.9498826   0.00692863]\n",
      " [20.89895873 17.20398296  4.03444117  0.11709897]\n",
      " [44.75658628 33.35566708 12.52980106  1.95579986]\n",
      " [36.98820168 35.39428509 22.58393674  1.80717374]]\n",
      "_______________________________________________\n",
      "Expected units sold per arm\n",
      "[[2.01934267 2.01861421 1.94987202 1.00860449]\n",
      " [1.         1.         1.         0.6       ]\n",
      " [2.97522029 3.04615342 3.00197665 3.33939304]\n",
      " [2.99704509 2.98706234 3.0092995  2.86256317]\n",
      " [1.         1.         1.         1.        ]]\n",
      "_______________________________________________\n",
      "Best configuration [0. 0. 0. 0. 0.]\n",
      "Optimal cumulative expected reward per round\n",
      "118.17314762734482\n"
     ]
    }
   ],
   "source": [
    "#FIND OPTIMAL SOLUTION AND CONVERSION RATES considering at the same time the three classes\n",
    "\n",
    "E = E_commerce(binary_features=1) #NORMAL ECOMMERCE\n",
    "E.set_lambda(0.5)\n",
    "E.set_products(products)\n",
    "E.graph = np.array([[0. , 0.5, 0. , 1. , 0. ],[1. , 0. , 0. , 0.5, 0. ],[0. , 1. , 0. , 0.5, 0. ],[0. , 0.5, 1. , 0. , 0. ],[1. , 0. , 0. , 0.5, 0. ]])\n",
    "\n",
    "\n",
    "n_arms = 4\n",
    "num_users = 100\n",
    "n_days = 365\n",
    "fixed_alpha = 0\n",
    "fixed_weights = 1\n",
    "fixed_units = 0\n",
    "num_experiments = 5\n",
    "opt_vector = np.zeros(num_experiments)\n",
    "conv_rates_per_experiment = []\n",
    "mean_units_sold_per_product_per_eperiment = []\n",
    "cr_learner_expected_rewards_per_experiment = []\n",
    "\n",
    "for e in range(num_experiments):\n",
    "    env = Environment(n_arms, E, margins_matrix, num_users, fixed_alpha, fixed_weights,fixed_units)\n",
    "    cr_learner = Conv_rates(n_arms=n_arms)\n",
    "\n",
    "    for d in range(n_days):\n",
    "\n",
    "        pulled_arm = cr_learner.pull_arm()#best_arm_per_product=np.array([2., 1., 1., 1., 0.]), clairvoyant_flag=1)\n",
    "        reward = env.round(pulled_arm)\n",
    "        cr_learner.update(pulled_arm, reward, env.clicks_current_day, env.purchases_current_day,env.daily_units)\n",
    "\n",
    "\n",
    "    conversion_rates = np.zeros((5,4))\n",
    "    np.set_printoptions(suppress=True)\n",
    "    for i in range(5):\n",
    "        conversion_rates[i] = cr_learner.beta_parameters[i][:,0]/(cr_learner.beta_parameters[i][:,0]+cr_learner.beta_parameters[i][:,1])\n",
    "        opt_vector[e] += np.max(np.array(E.products[i].margins_list) * conversion_rates[i] * cr_learner.lambda_poisson[i])\n",
    "\n",
    "    conv_rates_per_experiment.append(conversion_rates)\n",
    "    mean_units_sold_per_product_per_eperiment.append(cr_learner.lambda_poisson)\n",
    "    cr_learner_expected_rewards_per_experiment.append(cr_learner.expected_rewards)\n",
    "\n",
    "#optimal expected clarvoyant solution is given chosing each round the best combination\n",
    "opt = np.mean(opt_vector) #+ np.std(opt_vector)\n",
    "best_arm_per_product = np.zeros(5)\n",
    "for i in range(5):   \n",
    "    best_arm_per_product[i] = np.argmax(np.array(E.products[i].margins_list) * np.mean(conv_rates_per_experiment,axis=0)[i] \n",
    "                                        * np.mean(mean_units_sold_per_product_per_eperiment, axis = 0)[i])#expected_units_sold_per_product[i])#\n",
    "    \n",
    "print(\"_______________________________________________\")\n",
    "print(\"Conversion rates\")\n",
    "print(np.mean(conv_rates_per_experiment,axis=0))\n",
    "print(\"_______________________________________________\")\n",
    "print(\"Expected rewards per arm\")\n",
    "print(np.mean(cr_learner_expected_rewards_per_experiment,axis=0))\n",
    "print(\"_______________________________________________\")\n",
    "print(\"Expected units sold per arm\")\n",
    "print(np.mean(mean_units_sold_per_product_per_eperiment, axis = 0))\n",
    "\n",
    "print(\"_______________________________________________\")\n",
    "print(\"Best configuration\", best_arm_per_product)\n",
    "print(\"Optimal cumulative expected reward per round\")\n",
    "print(opt) # optimal configuration: the best combination of arms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 287,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([0., 0., 0., 0., 0.]), 118.17314762734482)"
      ]
     },
     "execution_count": 287,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "opt_3_classes = opt\n",
    "best_arm_per_product,opt_3_classes"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Estimate Conversion Rates considering the three classes one at time (with 3 conv_rates learners)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 288,
   "metadata": {},
   "outputs": [],
   "source": [
    "#FIND OPTIMAL SOLUTION AND CONVERSION RATES considering the classes separately\n",
    "\n",
    "#Class 3\n",
    "E = E_commerce(binary_features=30)\n",
    "E.set_lambda(0.5)\n",
    "E.set_products(products)\n",
    "E.graph = np.array([[0. , 0.5, 0. , 1. , 0. ],[1. , 0. , 0. , 0.5, 0. ],[0. , 1. , 0. , 0.5, 0. ],[0. , 0.5, 1. , 0. , 0. ],[1. , 0. , 0. , 0.5, 0. ]])\n",
    "\n",
    "n_arms = 4\n",
    "num_users = 100\n",
    "n_days = 365\n",
    "fixed_alpha = 0\n",
    "fixed_weights = 1\n",
    "fixed_units = 0\n",
    "num_experiments = 10\n",
    "opt_vector = np.zeros(num_experiments)\n",
    "conv_rates_per_experiment = []\n",
    "mean_units_sold_per_product_per_eperiment = []\n",
    "cr_learner_expected_rewards_per_experiment = []\n",
    "\n",
    "for e in range(num_experiments):\n",
    "    env = Environment(n_arms, E, margins_matrix, num_users, fixed_alpha, fixed_weights,fixed_units)\n",
    "    cr_learner = Conv_rates(n_arms=n_arms)\n",
    "\n",
    "    for d in range(n_days):\n",
    "\n",
    "        pulled_arm = cr_learner.pull_arm()#best_arm_per_product=np.array([2., 1., 1., 1., 0.]), clairvoyant_flag=1)\n",
    "        reward = env.round(pulled_arm)\n",
    "        cr_learner.update(pulled_arm, reward, env.clicks_current_day, env.purchases_current_day,env.daily_units)\n",
    "\n",
    "\n",
    "    conversion_rates = np.zeros((5,4))\n",
    "    np.set_printoptions(suppress=True)\n",
    "    for i in range(5):\n",
    "        conversion_rates[i] = cr_learner.beta_parameters[i][:,0]/(cr_learner.beta_parameters[i][:,0]+cr_learner.beta_parameters[i][:,1])\n",
    "        opt_vector[e] += np.max(np.array(E.products[i].margins_list) * conversion_rates[i] * cr_learner.lambda_poisson[i])\n",
    "\n",
    "    conv_rates_per_experiment.append(conversion_rates)\n",
    "    mean_units_sold_per_product_per_eperiment.append(cr_learner.lambda_poisson)\n",
    "    cr_learner_expected_rewards_per_experiment.append(cr_learner.expected_rewards)\n",
    "\n",
    "#optimal expected clarvoyant solution is given chosing each round the best combination\n",
    "opt3 = np.mean(opt_vector)\n",
    "best_arm_per_product3 = np.zeros(5)\n",
    "for i in range(5):   \n",
    "    best_arm_per_product3[i] = np.argmax(np.array(E.products[i].margins_list) * np.mean(conv_rates_per_experiment,axis=0)[i] \n",
    "                                        * np.mean(mean_units_sold_per_product_per_eperiment, axis = 0)[i])#expected_units_sold_per_product[i])#\n",
    "\n",
    "\n",
    "\n",
    "#Class 2\n",
    "E = E_commerce(binary_features=20)\n",
    "E.set_lambda(0.5)\n",
    "E.set_products(products)\n",
    "E.graph = np.array([[0. , 0.5, 0. , 1. , 0. ],[1. , 0. , 0. , 0.5, 0. ],[0. , 1. , 0. , 0.5, 0. ],[0. , 0.5, 1. , 0. , 0. ],[1. , 0. , 0. , 0.5, 0. ]])\n",
    "\n",
    "n_arms = 4\n",
    "num_users = 100\n",
    "n_days = 365\n",
    "fixed_alpha = 0\n",
    "fixed_weights = 1\n",
    "fixed_units = 0\n",
    "num_experiments = 10\n",
    "opt_vector = np.zeros(num_experiments)\n",
    "conv_rates_per_experiment = []\n",
    "mean_units_sold_per_product_per_eperiment = []\n",
    "cr_learner_expected_rewards_per_experiment = []\n",
    "\n",
    "for e in range(num_experiments):\n",
    "    env = Environment(n_arms, E, margins_matrix, num_users, fixed_alpha, fixed_weights,fixed_units)\n",
    "    cr_learner = Conv_rates(n_arms=n_arms)\n",
    "\n",
    "    for d in range(n_days):\n",
    "\n",
    "        pulled_arm = cr_learner.pull_arm()#best_arm_per_product=np.array([2., 1., 1., 1., 0.]), clairvoyant_flag=1)\n",
    "        reward = env.round(pulled_arm)\n",
    "        cr_learner.update(pulled_arm, reward, env.clicks_current_day, env.purchases_current_day,env.daily_units)\n",
    "\n",
    "\n",
    "    conversion_rates = np.zeros((5,4))\n",
    "    np.set_printoptions(suppress=True)\n",
    "    for i in range(5):\n",
    "        conversion_rates[i] = cr_learner.beta_parameters[i][:,0]/(cr_learner.beta_parameters[i][:,0]+cr_learner.beta_parameters[i][:,1])\n",
    "        opt_vector[e] += np.max(np.array(E.products[i].margins_list) * conversion_rates[i] * cr_learner.lambda_poisson[i])\n",
    "\n",
    "    conv_rates_per_experiment.append(conversion_rates)\n",
    "    mean_units_sold_per_product_per_eperiment.append(cr_learner.lambda_poisson)\n",
    "    cr_learner_expected_rewards_per_experiment.append(cr_learner.expected_rewards)\n",
    "\n",
    "#optimal expected clarvoyant solution is given chosing each round the best combination\n",
    "opt2 = np.mean(opt_vector) #+ np.std(opt_vector)\n",
    "best_arm_per_product2 = np.zeros(5)\n",
    "for i in range(5):   \n",
    "    best_arm_per_product2[i] = np.argmax(np.array(E.products[i].margins_list) * np.mean(conv_rates_per_experiment,axis=0)[i] \n",
    "                                        * np.mean(mean_units_sold_per_product_per_eperiment, axis = 0)[i])#expected_units_sold_per_product[i])#\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "#Class 1\n",
    "E = E_commerce(binary_features=10)\n",
    "E.set_lambda(0.5)\n",
    "E.set_products(products)\n",
    "E.graph = np.array([[0. , 0.5, 0. , 1. , 0. ],[1. , 0. , 0. , 0.5, 0. ],[0. , 1. , 0. , 0.5, 0. ],[0. , 0.5, 1. , 0. , 0. ],[1. , 0. , 0. , 0.5, 0. ]])\n",
    "\n",
    "n_arms = 4\n",
    "num_users = 100\n",
    "n_days = 365\n",
    "fixed_alpha = 0\n",
    "fixed_weights = 1\n",
    "fixed_units = 0\n",
    "num_experiments = 10\n",
    "opt_vector = np.zeros(num_experiments)\n",
    "conv_rates_per_experiment = []\n",
    "mean_units_sold_per_product_per_eperiment = []\n",
    "cr_learner_expected_rewards_per_experiment = []\n",
    "\n",
    "for e in range(num_experiments):\n",
    "    env = Environment(n_arms, E, margins_matrix, num_users, fixed_alpha, fixed_weights,fixed_units)\n",
    "    cr_learner = Conv_rates(n_arms=n_arms)\n",
    "\n",
    "    for d in range(n_days):\n",
    "\n",
    "        pulled_arm = cr_learner.pull_arm()#best_arm_per_product=np.array([2., 1., 1., 1., 0.]), clairvoyant_flag=1)\n",
    "        reward = env.round(pulled_arm)\n",
    "        cr_learner.update(pulled_arm, reward, env.clicks_current_day, env.purchases_current_day,env.daily_units)\n",
    "\n",
    "\n",
    "    conversion_rates = np.zeros((5,4))\n",
    "    np.set_printoptions(suppress=True)\n",
    "    for i in range(5):\n",
    "        conversion_rates[i] = cr_learner.beta_parameters[i][:,0]/(cr_learner.beta_parameters[i][:,0]+cr_learner.beta_parameters[i][:,1])\n",
    "        opt_vector[e] += np.max(np.array(E.products[i].margins_list) * conversion_rates[i] * cr_learner.lambda_poisson[i])\n",
    "\n",
    "    conv_rates_per_experiment.append(conversion_rates)\n",
    "    mean_units_sold_per_product_per_eperiment.append(cr_learner.lambda_poisson)\n",
    "    cr_learner_expected_rewards_per_experiment.append(cr_learner.expected_rewards)\n",
    "\n",
    "#optimal expected clarvoyant solution is given chosing each round the best combination\n",
    "opt1 = np.mean(opt_vector) #+ np.std(opt_vector)\n",
    "best_arm_per_product1 = np.zeros(5)\n",
    "for i in range(5):   \n",
    "    best_arm_per_product1[i] = np.argmax(np.array(E.products[i].margins_list) * np.mean(conv_rates_per_experiment,axis=0)[i] \n",
    "                                        * np.mean(mean_units_sold_per_product_per_eperiment, axis = 0)[i])#expected_units_sold_per_product[i])#\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 289,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(144.5350815999111, 122.37979173130459, 111.18564330680545)"
      ]
     },
     "execution_count": 289,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "opt1,opt2,opt3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 290,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "125.66813416453681"
      ]
     },
     "execution_count": 290,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "opt_one_at_time = 0.3*opt1 + 0.4*opt2+0.3*opt3\n",
    "opt_one_at_time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 291,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([1., 1., 1., 1., 2.]),\n",
       " array([0., 0., 0., 0., 1.]),\n",
       " array([0., 0., 0., 0., 0.]))"
      ]
     },
     "execution_count": 291,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_arm_per_product1,best_arm_per_product2,best_arm_per_product3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "TS considering the first split, Class1 + Class2 (two learners, one class at time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 292,
   "metadata": {},
   "outputs": [],
   "source": [
    "#TS First split, 2 classes\n",
    "\n",
    "\n",
    "E = E_commerce(binary_features=20)\n",
    "E.set_lambda(0.5)\n",
    "E.set_products(products)\n",
    "E.graph = np.array([[0. , 0.5, 0. , 1. , 0. ],[1. , 0. , 0. , 0.5, 0. ],[0. , 1. , 0. , 0.5, 0. ],[0. , 0.5, 1. , 0. , 0. ],[1. , 0. , 0. , 0.5, 0. ]])\n",
    "\n",
    "n_arms = 4\n",
    "num_users = 70*3\n",
    "n_days = 365\n",
    "fixed_alpha = 0\n",
    "fixed_weights = 1\n",
    "fixed_units = 0\n",
    "n_experiments = 10\n",
    "\n",
    "ts2_rewards_per_experiment = []\n",
    "\n",
    "\n",
    "ts2_pulls_per_arm_per_experiment = []\n",
    "\n",
    "\n",
    "\n",
    "for e in range(n_experiments):\n",
    "\n",
    "    env = Environment(n_arms, E, margins_matrix, num_users, fixed_alpha,fixed_weights,fixed_units)\n",
    "    ts_learner = TS_poisson(n_arms=n_arms)\n",
    "\n",
    "    for d in range(n_days):\n",
    "        \n",
    "        pulled_arm = ts_learner.pull_arm(env.margins_matrix)\n",
    "        reward = env.round(pulled_arm)\n",
    "        ts_learner.update(pulled_arm, reward, env.clicks_current_day, env.purchases_current_day, env.daily_units)\n",
    "\n",
    "    ts2_rewards_per_experiment.append(ts_learner.collected_rewards)\n",
    "\n",
    "    ts2_pulls_per_arm_per_experiment.append(ts_learner.counter_per_arm)\n",
    "\n",
    "\n",
    "\n",
    "E = E_commerce(binary_features=10)\n",
    "E.set_lambda(0.5)\n",
    "E.set_products(products)\n",
    "E.graph = np.array([[0. , 0.5, 0. , 1. , 0. ],[1. , 0. , 0. , 0.5, 0. ],[0. , 1. , 0. , 0.5, 0. ],[0. , 0.5, 1. , 0. , 0. ],[1. , 0. , 0. , 0.5, 0. ]])\n",
    "\n",
    "n_arms = 4\n",
    "num_users = 30*3\n",
    "n_days = 365\n",
    "fixed_alpha = 0\n",
    "fixed_weights = 1\n",
    "fixed_units = 0\n",
    "n_experiments = 10\n",
    "\n",
    "ts1_rewards_per_experiment = []\n",
    "\n",
    "\n",
    "ts1_pulls_per_arm_per_experiment = []\n",
    "\n",
    "\n",
    "for e in range(n_experiments):\n",
    "\n",
    "    env = Environment(n_arms, E, margins_matrix, num_users, fixed_alpha,fixed_weights,fixed_units)\n",
    "    ts_learner = TS_poisson(n_arms=n_arms)\n",
    "\n",
    "    for d in range(n_days):\n",
    "        \n",
    "        pulled_arm = ts_learner.pull_arm(env.margins_matrix)\n",
    "        reward = env.round(pulled_arm)\n",
    "        ts_learner.update(pulled_arm, reward, env.clicks_current_day, env.purchases_current_day, env.daily_units)\n",
    "\n",
    "    ts1_rewards_per_experiment.append(ts_learner.collected_rewards)\n",
    "\n",
    "    ts1_pulls_per_arm_per_experiment.append(ts_learner.counter_per_arm)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 293,
   "metadata": {},
   "outputs": [],
   "source": [
    "ts_reward_first_split = 0.3*(np.mean(ts1_rewards_per_experiment, axis=0))+0.7*(np.mean(ts2_rewards_per_experiment, axis=0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 294,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "129.02637869188655"
      ]
     },
     "execution_count": 294,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "opt12 = 0.3*opt1+0.7*opt2\n",
    "opt12"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 295,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(0)\n",
    "plt.xlabel(\"t\")\n",
    "plt.ylabel(\"Regret\")\n",
    "plt.title(\"1st and 2nd class with two learners, first split\")\n",
    "plt.plot(np.cumsum(opt12-ts_reward_first_split, axis=0), 'r')\n",
    "plt.legend([\"TS\"])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "TS considering the second split, 3 classes (3 learners, one class at time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 296,
   "metadata": {},
   "outputs": [],
   "source": [
    "#TS class 0 + 1 + 2 for tot days (3 classes, one at time)\n",
    "\n",
    "E = E_commerce(binary_features=30)\n",
    "E.set_lambda(0.5)\n",
    "E.set_products(products)\n",
    "E.graph = np.array([[0. , 0.5, 0. , 1. , 0. ],[1. , 0. , 0. , 0.5, 0. ],[0. , 1. , 0. , 0.5, 0. ],[0. , 0.5, 1. , 0. , 0. ],[1. , 0. , 0. , 0.5, 0. ]])\n",
    "\n",
    "n_arms = 4\n",
    "num_users = 30*3\n",
    "n_days = 365\n",
    "fixed_alpha = 0\n",
    "fixed_weights = 1\n",
    "fixed_units = 0\n",
    "n_experiments = 10\n",
    "\n",
    "ts3_rewards_per_experiment = []\n",
    "\n",
    "\n",
    "ts3_pulls_per_arm_per_experiment = []\n",
    "\n",
    "\n",
    "\n",
    "for e in range(n_experiments):\n",
    "\n",
    "    env = Environment(n_arms, E, margins_matrix, num_users, fixed_alpha,fixed_weights,fixed_units)\n",
    "    ts_learner = TS_poisson(n_arms=n_arms)\n",
    "\n",
    "    for d in range(n_days):\n",
    "        \n",
    "        pulled_arm = ts_learner.pull_arm(env.margins_matrix)\n",
    "        reward = env.round(pulled_arm)\n",
    "        ts_learner.update(pulled_arm, reward, env.clicks_current_day, env.purchases_current_day, env.daily_units)\n",
    "\n",
    "    ts3_rewards_per_experiment.append(ts_learner.collected_rewards)\n",
    "\n",
    "    ts3_pulls_per_arm_per_experiment.append(ts_learner.counter_per_arm)\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "E = E_commerce(binary_features=20)\n",
    "E.set_lambda(0.5)\n",
    "E.set_products(products)\n",
    "E.graph = np.array([[0. , 0.5, 0. , 1. , 0. ],[1. , 0. , 0. , 0.5, 0. ],[0. , 1. , 0. , 0.5, 0. ],[0. , 0.5, 1. , 0. , 0. ],[1. , 0. , 0. , 0.5, 0. ]])\n",
    "\n",
    "n_arms = 4\n",
    "num_users = 40*3\n",
    "n_days = 365\n",
    "fixed_alpha = 0\n",
    "fixed_weights = 1\n",
    "fixed_units = 0\n",
    "n_experiments = 10\n",
    "\n",
    "ts2_rewards_per_experiment = []\n",
    "\n",
    "\n",
    "ts2_pulls_per_arm_per_experiment = []\n",
    "\n",
    "\n",
    "\n",
    "for e in range(n_experiments):\n",
    "\n",
    "    env = Environment(n_arms, E, margins_matrix, num_users, fixed_alpha,fixed_weights,fixed_units)\n",
    "    ts_learner = TS_poisson(n_arms=n_arms)\n",
    "\n",
    "    for d in range(n_days):\n",
    "        \n",
    "        pulled_arm = ts_learner.pull_arm(env.margins_matrix)\n",
    "        reward = env.round(pulled_arm)\n",
    "        ts_learner.update(pulled_arm, reward, env.clicks_current_day, env.purchases_current_day, env.daily_units)\n",
    "\n",
    "    ts2_rewards_per_experiment.append(ts_learner.collected_rewards)\n",
    "\n",
    "    ts2_pulls_per_arm_per_experiment.append(ts_learner.counter_per_arm)\n",
    "\n",
    "\n",
    "\n",
    "E = E_commerce(binary_features=10)\n",
    "E.set_lambda(0.5)\n",
    "E.set_products(products)\n",
    "E.graph = np.array([[0. , 0.5, 0. , 1. , 0. ],[1. , 0. , 0. , 0.5, 0. ],[0. , 1. , 0. , 0.5, 0. ],[0. , 0.5, 1. , 0. , 0. ],[1. , 0. , 0. , 0.5, 0. ]])\n",
    "\n",
    "n_arms = 4\n",
    "num_users = 30*3\n",
    "n_days = 365\n",
    "fixed_alpha = 0\n",
    "fixed_weights = 1\n",
    "fixed_units = 0\n",
    "n_experiments = 10\n",
    "\n",
    "ts1_rewards_per_experiment = []\n",
    "\n",
    "\n",
    "ts1_pulls_per_arm_per_experiment = []\n",
    "\n",
    "\n",
    "for e in range(n_experiments):\n",
    "\n",
    "    env = Environment(n_arms, E, margins_matrix, num_users, fixed_alpha,fixed_weights,fixed_units)\n",
    "    ts_learner = TS_poisson(n_arms=n_arms)\n",
    "\n",
    "    for d in range(n_days):\n",
    "        \n",
    "        pulled_arm = ts_learner.pull_arm(env.margins_matrix)\n",
    "        reward = env.round(pulled_arm)\n",
    "        ts_learner.update(pulled_arm, reward, env.clicks_current_day, env.purchases_current_day, env.daily_units)\n",
    "\n",
    "    ts1_rewards_per_experiment.append(ts_learner.collected_rewards)\n",
    "\n",
    "    ts1_pulls_per_arm_per_experiment.append(ts_learner.counter_per_arm)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 297,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TS 141.43407433789957 14.13719457919283\n",
      "TS 120.46212729223743 12.205016210565221\n",
      "TS 107.24539356621005 11.351312159602294\n"
     ]
    }
   ],
   "source": [
    "print(\"TS\",np.mean(np.mean(ts1_rewards_per_experiment, axis=0)), np.std(np.mean(ts1_rewards_per_experiment, axis=0)))\n",
    "print(\"TS\",np.mean(np.mean(ts2_rewards_per_experiment, axis=0)), np.std(np.mean(ts2_rewards_per_experiment, axis=0)))\n",
    "print(\"TS\",np.mean(np.mean(ts3_rewards_per_experiment, axis=0)), np.std(np.mean(ts3_rewards_per_experiment, axis=0)))\n",
    "\n",
    "ts_all_classes_rewards = 0.3*(np.mean(ts1_rewards_per_experiment, axis=0))+0.4*(np.mean(ts2_rewards_per_experiment, axis=0))+0.3*np.mean(ts3_rewards_per_experiment, axis=0)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 298,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "125.66813416453681"
      ]
     },
     "execution_count": 298,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "opt123 = 0.3*opt1+0.4*opt2+0.3*opt3\n",
    "opt123"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 299,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(0)\n",
    "plt.xlabel(\"t\")\n",
    "plt.ylabel(\"Regret\")\n",
    "plt.title(\"1st class\")\n",
    "plt.plot(np.cumsum(opt1-np.mean(ts1_rewards_per_experiment,axis=0), axis=0), 'r')\n",
    "plt.legend([\"TS\"])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 300,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(0)\n",
    "plt.xlabel(\"t\")\n",
    "plt.ylabel(\"Regret\")\n",
    "plt.title(\"2nd class\")\n",
    "plt.plot(np.cumsum(opt2-np.mean(ts2_rewards_per_experiment,axis=0), axis=0), 'r')\n",
    "plt.legend([\"TS\"])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 301,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(0)\n",
    "plt.xlabel(\"t\")\n",
    "plt.ylabel(\"Regret\")\n",
    "plt.title(\"3rd class\")\n",
    "plt.plot(np.cumsum(opt3-np.mean(ts3_rewards_per_experiment,axis=0), axis=0), 'r')\n",
    "plt.legend([\"TS\"])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 302,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAkQAAAHFCAYAAAAT5Oa6AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/MnkTPAAAACXBIWXMAAA9hAAAPYQGoP6dpAABYVklEQVR4nO3deVhUZf8G8HvYhkUYQZEBRSR3Ra3cckXFXTRLzdQUl0p/LkVquaZWCmqllppLuZBammuLSy4Zar4uoSaimSbuIC4wILvw/P54YnAEFHHgDDP357rmmpkzZ858z4zvy91znkUlhBAgIiIismBWShdAREREpDQGIiIiIrJ4DERERERk8RiIiIiIyOIxEBEREZHFYyAiIiIii8dARERERBaPgYiIiIgsHgMRERERWTwGIqJi9OWXX0KlUsHPz6/AfVQqFWbMmKF//vvvv0OlUuH3339/5s+/fPkyVCoVVq9e/czHMjdt2rRBmzZt9M9TUlIwY8aMfL/3GTNmQKVS4c6dO0X6rO+//x6tW7eGh4cH1Go1vLy80L17dxw+fLjQtT7u3xARPTsbpQsgMmcrV64EAERFReHo0aNo2rSpwhVRjq+++srgeUpKCj766CMAMAhKxnD37l20aNEC7777LsqXL4+YmBjMmzcPrVu3xr59++Dv72/UzyOip8dARFRM/vzzT/z111/o1q0btm/fjhUrVjAQmZA6deqU2GeNHj06z7YuXbrA3d0dK1asKDWBKDU1Ffb29lCpVMX+WUIIpKWlwcHBodg/iwjgJTOiYrNixQoAwOzZs9G8eXOsX78eKSkpRv2MGzdu4O2334a3tzfs7Ozg5eWF3r1749atWwW+5+LFixgyZAiqV68OR0dHVKxYEd27d0dkZKTBftnZ2Zg5cyZq1qwJBwcHlC1bFvXr18cXX3yh3+f27dv6z1er1XB3d0eLFi2wd+9eg2Pt3bsXAQEBcHFxgaOjI1q0aIF9+/YZ7FPYYz0sKioKKpUKGzdu1G+LiIiASqVC3bp1Dfbt0aMHGjZsqH/+8CWzy5cvw93dHQDw0UcfQaVSQaVSYfDgwQbHuHXrFvr16weNRgMPDw8MHToUOp2uwPoex9nZGfb29rCxMd5/l27YsAHNmjWDk5MTypQpg06dOuHkyZMG+/z55594/fXXUaVKFTg4OKBKlSro168frly5YrDf6tWroVKpsHv3bgwdOhTu7u5wdHREenq6/hLe8ePH0apVKzg6OuK5557D7NmzkZ2dbXCcxMREjB8/Hr6+vrCzs0PFihURHByM5ORkg/1UKhVGjx6NpUuXonbt2lCr1QgLCwMALFmyBA0aNECZMmXg7OyMWrVqYfLkyUb73ogAthARFYvU1FR8//33aNy4Mfz8/DB06FC8+eab2LhxI4KCgozyGTdu3EDjxo2RmZmJyZMno379+rh79y5+/fVXxMfHw8PDI9/33bx5E+XKlcPs2bPh7u6Oe/fuISwsDE2bNsXJkydRs2ZNAMDcuXMxY8YMTJ06Fa1bt0ZmZib+/vtvJCQk6I81cOBAnDhxArNmzUKNGjWQkJCAEydO4O7du/p91q5di0GDBuHll19GWFgYbG1tsWzZMnTq1Am//vorAgICCn2sR9WtWxeenp7Yu3cv+vTpA0CGLwcHB5w9exY3b96El5cXHjx4gPDwcIwYMSLf43h6emLXrl3o3Lkzhg0bhjfffBMA9CEpR69evdC3b18MGzYMkZGRmDRpEoDcS6NPkpWVhezsbNy4cQOhoaEQQmDUqFGFeu+ThISEYOrUqRgyZAimTp2KjIwMfPrpp2jVqhWOHTumbxG7fPkyatasiddffx1ubm6IiYnBkiVL0LhxY5w9exbly5c3OO7QoUPRrVs3rFmzBsnJybC1tQUAxMbGYsCAARg3bhymT5+OrVu3YtKkSfDy8sKgQYMAyMuQ/v7+uH79uv7faFRUFKZNm4bIyEjs3bvXoLVp27ZtOHjwIKZNmwatVosKFSpg/fr1GDlyJMaMGYPPPvsMVlZWuHjxIs6ePWuU741ITxCR0X377bcCgFi6dKkQQoikpCRRpkwZ0apVqzz7AhDTp0/XP9+/f78AIPbv3//Yzxg6dKiwtbUVZ8+eLXCf6OhoAUCsWrWqwH0ePHggMjIyRPXq1cV7772n3x4YGCief/75x9ZQpkwZERwcXODrycnJws3NTXTv3t1ge1ZWlmjQoIFo0qRJoY9VkDfeeEM899xz+uft27cXb731lnB1dRVhYWFCCCH++OMPAUDs3r1bv5+/v7/w9/fXP799+3ae3yLH9OnTBQAxd+5cg+0jR44U9vb2Ijs7u1C11qxZUwAQAISnp6c4dOhQod7n7+8v6tatW+DrV69eFTY2NmLMmDEG25OSkoRWqxWvvfZage998OCBuH//vnBychJffPGFfvuqVasEADFo0KB86wEgjh49arC9Tp06olOnTvrnoaGhwsrKShw/ftxgv02bNgkAYseOHfptAIRGoxH37t0z2Hf06NGibNmyBdZPZCy8ZEZUDFasWAEHBwe8/vrrAIAyZcqgT58+OHjwIC5cuGCUz9i5cyfatm2L2rVrP9X7Hjx4gJCQENSpUwd2dnawsbGBnZ0dLly4gHPnzun3a9KkCf766y+MHDkSv/76KxITE/Mcq0mTJli9ejVmzpyJI0eOIDMz0+D1w4cP4969ewgKCsKDBw/0t+zsbHTu3BnHjx/XXzp50rEKEhAQgEuXLiE6OhppaWk4dOgQOnfujLZt22LPnj0AZKuRWq1Gy5Ytn+q7elSPHj0MntevXx9paWmIi4sr1Ps3b96Mo0ePYuPGjahTpw66dOlilNGEv/76Kx48eIBBgwYZfM/29vbw9/c3+Iz79+9jwoQJqFatGmxsbGBjY4MyZcogOTnZ4PfP0atXr3w/U6vVokmTJgbb6tevb3Dp7ZdffoGfnx+ef/55g7o6deqU70jKdu3awdXV1WBbkyZNkJCQgH79+uHHH38s8kg/oidhICIysosXL+LAgQPo1q0bhBBISEhAQkICevfuDaDwl1ee5Pbt26hUqdJTv2/s2LH48MMP0bNnT/z88884evQojh8/jgYNGiA1NVW/36RJk/DZZ5/hyJEj6NKlC8qVK4eAgAD8+eef+n02bNiAoKAgfPPNN2jWrBnc3NwwaNAgxMbGAoC+L1Pv3r1ha2trcJszZw6EELh3716hjlWQ9u3bA5Ch59ChQ8jMzES7du3Qvn17fT+lvXv3okWLFs/cQbdcuXIGz9VqNQAYfG+PU7duXTRp0gS9e/fGrl274OPjg3ffffeZagJyv+fGjRvn+Z43bNhgECL69++PRYsW4c0338Svv/6KY8eO4fjx43B3d8/3PDw9PfP9zEe/C0B+Hw8f49atWzh9+nSempydnSGEyBNu8vusgQMHYuXKlbhy5Qp69eqFChUqoGnTpvqwS2Qs7ENEZGQrV66EEAKbNm3Cpk2b8rweFhaGmTNnwtra+pk+x93dHdevX3/q9+X06QkJCTHYfufOHZQtW1b/3MbGBmPHjsXYsWORkJCAvXv3YvLkyejUqROuXbsGR0dHlC9fHgsWLMCCBQtw9epV/PTTT5g4cSLi4uKwa9cufX+UhQsX4qWXXsq3npy+Tk86VkEqVaqEGjVqYO/evahSpQoaNWqEsmXLIiAgACNHjsTRo0dx5MgR/ZB6U2FjY4MXX3wRP/zwwzMfK+d73rRpE3x8fArcT6fT4ZdffsH06dMxceJE/fb09HR9MH3Us4woK1++PBwcHAr8j4BH+ysV9FlDhgzBkCFDkJycjAMHDmD69OkIDAzEP//889jzJXoaDERERpSVlYWwsDBUrVoV33zzTZ7Xf/nlF3z++efYuXMnAgMDn+mzunTpgjVr1uD8+fP6jtCFoVKp9C0bObZv344bN26gWrVq+b6nbNmy6N27N27cuIHg4GBcvnw5z7D1ypUrY/To0di3bx/++OMPAECLFi1QtmxZnD17Nt+h5wXJ71iP0759e/zwww/w9vZGt27dAAA1atRA5cqVMW3aNGRmZupbkgrytK09zyotLQ1Hjhwp8Dt/Gp06dYKNjQ3+/fffAi9xAfK3F0Lk+f2/+eYbZGVlPXMdjwoMDERISAjKlSsHX1/fZz6ek5MTunTpgoyMDPTs2RNRUVEMRGQ0DERERrRz507cvHkTc+bMyXdyPz8/PyxatAgrVqx45kD08ccfY+fOnWjdujUmT56MevXqISEhAbt27cLYsWNRq1atfN8XGBiI1atXo1atWqhfvz4iIiLw6aef5rn81r17d/j5+aFRo0Zwd3fHlStXsGDBAvj4+KB69erQ6XRo27Yt+vfvj1q1asHZ2RnHjx/Hrl278OqrrwKQfacWLlyIoKAg3Lt3D71790aFChVw+/Zt/PXXX7h9+zaWLFlSqGM9TkBAAL766ivcuXMHCxYsMNi+atUquLq6Ggy5z4+zszN8fHzw448/IiAgAG5ubihfvjyqVKnyxM9/kubNm6NHjx6oXbs2NBoNLl++jCVLluDff//F1q1bC3WMxMTEfFsc3d3d4e/vj48//hhTpkzBpUuX0LlzZ7i6uuLWrVs4duwYnJyc8NFHH8HFxQWtW7fGp59+qj+38PBwrFixwqB10FiCg4OxefNmtG7dGu+99x7q16+P7OxsXL16Fbt378a4ceOeODfXW2+9BQcHB7Ro0QKenp6IjY1FaGgoNBoNGjdubPSayYIp2qWbyMz07NlT2NnZibi4uAL3ef3114WNjY2IjY0VQhR9lJkQQly7dk0MHTpUaLVaYWtrK7y8vMRrr70mbt26JYTIf5RZfHy8GDZsmKhQoYJwdHQULVu2FAcPHswz6urzzz8XzZs3F+XLlxd2dnaicuXKYtiwYeLy5ctCCCHS0tLEiBEjRP369YWLi4twcHAQNWvWFNOnTxfJyckGdYaHh4tu3boJNzc3YWtrKypWrCi6desmNm7c+NTHyk98fLywsrISTk5OIiMjQ7993bp1AoB49dVX87zn0fMVQoi9e/eKF154QajVagFABAUFCSFyR5ndvn3bYP+ckVjR0dGPrW/cuHGiQYMGQqPRCBsbG6HVasUrr7wi/vjjjyeeW06t+G902qO3h89h27Ztom3btsLFxUWo1Wrh4+MjevfuLfbu3avf5/r166JXr17C1dVVODs7i86dO4szZ84IHx8f/fk+fG6PjhDLqSe/UW9BQUHCx8fHYNv9+/fF1KlTRc2aNYWdnZ3QaDSiXr164r333tP/b0AI+b+DUaNG5TlmWFiYaNu2rfDw8BB2dnb6f+OnT58u1HdHVFgqIYRQIIcRERERmQyOMiMiIiKLx0BEREREFo+BiIiIiCweAxERERFZPAYiIiIisngMRERERGTxODFjIWVnZ+PmzZtwdnZ+pqnsiYiIqOQIIZCUlAQvLy9YWRXcDsRAVEg3b96Et7e30mUQERFREVy7du2xC2IzEBWSs7MzAPmFuri4KFwNERERFUZiYiK8vb31f8cLwkBUSDmXyVxcXBiIiIiISpkndXdhp2oiIiKyeAxEREREZPEYiIiIiMjisQ+RkWVlZSEzM1PpMhRha2sLa2trpcsgIiJ6agxERiKEQGxsLBISEpQuRVFly5aFVqvlXE1ERFSqMBAZSU4YqlChAhwdHS0uEAghkJKSgri4OACAp6enwhUREREVHgOREWRlZenDULly5ZQuRzEODg4AgLi4OFSoUIGXz4iIqNRgp2ojyOkz5OjoqHAlysv5Diy1HxUREZVODERGZGmXyfLD74CIiEojBiIiIiKyeAxEREREZPEYiCyUSqV67G3w4MEAgP3796Nt27Zwc3ODo6MjqlevjqCgIDx48EDZEyAiIjIijjKzUDExMfrHGzZswLRp03D+/Hn9NgcHB0RFRaFLly545513sHDhQjg4OODChQvYtGkTsrOzlSibiIhKs+xsID0d+G9UsilhILJQWq1W/1ij0UClUhlsA4BVq1bB09MTc+fO1W+rWrUqOnfuXGJ1EhFRKZeaCnz/PRATA3zxBRAfD7RpA8yYAdy5I193dJS3F14AFJq+hoGoOAgBpKQo89mOjoCRRnpptVrExMTgwIEDaN26tVGOSUREZiopCTh9Gmje3PDvUGgo8Mknhvvu3Stvj9qxA+jSpXjrLAADUXFISQHKlFHms+/fB5ycjHKoPn364Ndff4W/vz+0Wi1eeuklBAQEYNCgQXBxcTHKZxARkZkYPRr49lvZ8jN9utyWmQl8/bV87OkJTJkCtG8PfPghsHEjULUqULmy/LuZkqJY6xDAQESPYW1tjVWrVmHmzJn47bffcOTIEcyaNQtz5szBsWPHuDwHERFJ9+/LMATIQNSqFdCuHfDLL0BsLODhAVy5Atjayn1++AFISwPUaqNd1XhWDETFwdFR/uNQ6rONrGLFihg4cCAGDhyImTNnokaNGli6dCk++ugjo38WERGVQj//bPi8c2dgxAhgyxb5fMiQ3DCUw96+ZGorJAai4qBSGe2ylalxdXWFp6cnkpOTlS6FiIgKY/duYOVK4NNPAW/v4vmMDRvk/XvvAdeuAZs2AQsXym01awLBwcXzuUbEQEQFWrZsGU6dOoVXXnkFVatWRVpaGr799ltERUVhYc4/dCIiMl0ZGcCbb8qQcucOsGeP8S9RxcQA27fLx0OHAnXrykC0Zw/g7g5MnAg4Oxv3M4sBAxEVqEmTJjh06BBGjBiBmzdvokyZMqhbty62bdsGf39/pcsjIqIn+f57GYYAYN8+YPVqefnKmJYtAx48AFq0APz85LY+feStFFEJIYTSRZQGiYmJ0Gg00Ol0eUZYpaWlITo6Gr6+vrA3sWuiJY3fBRGRCalfH4iMBOrVk/dVqgAXLgA2RmoPyciQo8Ru3ZLh6/XXjXNcI3rc3++HcekOIiIic3T9ugxBVlbArl3y8tXly3K4+6NSUoo2GCg8XIahChWAXr2euWQlMRARERGZo/Bwef/CC4CXFzBmjHw+d66cQBgA7t4Fhg+X8/+4uclJEd9+Wwanwti5U95365Z3FFkpwz5ERERE5ignELVpI+9HjQJmzwZOnZJD4i9dki1It27lvmfXLnkfGwv89JPh8YQAliwBkpOB7t2BWrVy91dodmljYgsRERGROUlNBRYtyp0hOicQubkBb70lHy9fLpfOuHVLDov//XcZlObPl6PQfv5ZLsORlAQcPCj7HY0dK0PVBx8AtWsD1asD584B1tZAhw4KnKhxsYXIiNg/nd8BEZFiMjOBxEQZWnLmBVKpgJYtc/d57z0ZlrKy5BD5V16RM0rnTOrboAHwv//JmaTffluOULt50/BzWrUCjh4FLl6Uz5s3B8qWLfbTK26KthAdOHAA3bt3h5eXF1QqFbZt22bwuhACM2bMgJeXFxwcHNCmTRtERUUZ7JOeno4xY8agfPnycHJyQo8ePXD9+nWDfeLj4zFw4EBoNBpoNBoMHDgQCQkJRjsP2/+um6YotaCrCcn5DmxL+bVkIqJSJS4OaNwYKF9ehiEbG6BaNWDkSMOw4uMjO1UvXgx88w0QGJh3hYNp0+S2o0dlGHJxkbNK16wpJ3g8cEDOabRxo2w1MpN56RRtIUpOTkaDBg0wZMgQ9Mqnd/rcuXMxb948rF69GjVq1MDMmTPRoUMHnD9/Hs7/TfIUHByMn3/+GevXr0e5cuUwbtw4BAYGIiIiAtbW1gCA/v374/r169j137XOt99+GwMHDsTPj041XkTW1tYoW7Ys4uLiAACOjo5QmcjaLCVFCIGUlBTExcWhbNmy+u+eiIiKUWam7BgdEACcPZu7fcEC2VKUn1deefwx69YFIiJkJ2wHByAsDHB1NdzH2Rno3VvezITJzEOkUqmwdetW9OzZE4D8A+vl5YXg4GBMmDABgGwN8vDwwJw5czB8+HDodDq4u7tjzZo16Nu3LwDg5s2b8Pb2xo4dO9CpUyecO3cOderUwZEjR9C0aVMAwJEjR9CsWTP8/fffqFmzZqHqe9I8BkIIxMbGGrXlqTQqW7YstFqtxQVCIqISt3ixXGHe2Vn29alYEdixQ44Yq1hR6epMRmHnITLZPkTR0dGIjY1Fx44d9dvUajX8/f1x+PBhDB8+HBEREcjMzDTYx8vLC35+fjh8+DA6deqE//3vf9BoNPowBAAvvfQSNBoNDh8+XGAgSk9PR3p6uv55YmLiY+tVqVTw9PREhQoVkJmZWdTTLtVsbW3ZMkREZGy3bsnOyzmdowE5IeLMmfJxUpJco2z/fqBqVUVKNAcmG4hiY2MBAB4eHgbbPTw8cOXKFf0+dnZ2cH2kKc/Dw0P//tjYWFSoUCHP8StUqKDfJz+hoaFFWs3d2tqaoYCIiIwjKwvw9wfOnwd++w1o21Zu37xZDo23tZXLcbRrB2i1ipZa2pn8sPtHL70IIZ54OebRffLb/0nHmTRpEnQ6nf52LWctGCIiopKybp0MQ4BcGgOQ/YbmzpWPp04F+vdnGDICk20h0v7348bGxsLT01O/PS4uTt9qpNVqkZGRgfj4eINWori4ODRv3ly/z62HJ536z+3bt/O0Pj1MrVZDrVYb5VyIiIieWmoqMGNG7vOffpKjv/79V84ZVLasnGCRjMJkW4h8fX2h1WqxZ88e/baMjAyEh4frw07Dhg1ha2trsE9MTAzOnDmj36dZs2bQ6XQ4duyYfp+jR49Cp9Pp9yEiIjI5n3wCREfLTtOA7Ev0xRfAL7/I50uWyDXEyCgUbSG6f/8+LuZM7ATZkfrUqVNwc3ND5cqVERwcjJCQEFSvXh3Vq1dHSEgIHB0d0b9/fwCARqPBsGHDMG7cOJQrVw5ubm4YP3486tWrh/bt2wMAateujc6dO+Ott97CsmXLAMhh94GBgYUeYUZERFSijhzJvSz27bfA5MmyYzUADBwoV683wZXlSzWhoP379wsAeW5BQUFCCCGys7PF9OnThVarFWq1WrRu3VpERkYaHCM1NVWMHj1auLm5CQcHBxEYGCiuXr1qsM/du3fFgAEDhLOzs3B2dhYDBgwQ8fHxT1WrTqcTAIROp3uWUyYiInq869eF0GqFAIR47TW57fvvhXBwEGL5cmVrK4UK+/fbZOYhMnWFnceAiIio0LKz5ZpijRvnTn44aBCwZo1sBTp8GChTJndfK5Pt6WKyCvv3m98sERFRSYmLAxo1AoKC5OrxU6YAnTrJ8LNwIbB+PbB2rdx35crcMAQwDBUzkx1lRkREZFaEkKPCIiLkrWJFYN48+dqNG8A77+Tu2769DE5UYhiIiIiIisPdu3JovLU18OCBbA3aujX39dBQed+iBdCxI3DsGPDHH0BysuFweyoRDERERETGtmgR8O67wJtvAsuWAR9+mDtqbO5c4NAhudTGc8/JS2M1asjXMjOBlBRAo1GudgvFTtWFxE7VRERUKGvXyqHxOc6eleuQxcXJuYM4mWKJYqdqIiKikpaYCIwda7itXTsZhlxdgWHDlKmLnoiBiIiIyFjmzgVu35aXwI4eldtyFhLv2VMuxkomiYGIiIjIGLKzZX8hAAgJAZo0AebPz329Tx9l6qJCYadqIiIiYzh1CrhzR84d1KOH3BYcLC+V/fOPnG+ITBYDERERkTHkLDTetq3hpbGgIGXqoafCS2ZERETGkBOIOnRQtg4qErYQERERPYv0dGDTJuDgQfmcgahUYgsRERFRUdy9K/sIlSsHvPEGkJEhh9jXrKl0ZVQEbCEiIiIqip495YzTAKDVAiNHAhMmACqVomVR0TAQERERFVZEBLBgAeDuLsOQnR3w88/yMhmDUKnGQERERPQkDx4AkyYBn31muL13b7kwK5V67ENERET0JG+9lRuG6tbN3f5//6dMPWR0bCEiIiJ6nL//BsLC5OMffpB9h0aNAuztgRYtFC2NjIeBiIiI6HHmzAGEkLNP5yy/sXy5sjWR0fGSGRERUUE2b85tHZoyRdlaqFgxEBEREeU4elSOGHN3ByZOlPMLCQGMHi0XayWzxUtmREREAHDjBhAQACQny+dz5sj7nj3lUHsya2whIiIiAuSkisnJQOPGwLhxgI0N8PbbwMaNgLW10tVRMVMJIYTSRZQGiYmJ0Gg00Ol0cHFxUbocIiIypogIoFEjObni8eNAw4ZyjTK1WunK6BkV9u83W4iIiIg++UTeDxggwxDAMGRhGIiIiMiy/fUX8OOPsnVo6lSlqyGFsFM1ERFZpuxsuUL9/Pny+WuvcaV6C8ZARERElkcIoF07IDw8d1twsGLlkPJ4yYyIiCzPwYOGYej554GmTRUrh5THQERERJZn4ULD5++/L/sQkcXiJTMiIrIsV68CW7fKx3/+KecbatBA2ZpIcWwhIiKi0i87G1i2DHjhBaBbN9lZuiDz5gFZWUDbtnKIPcMQgS1ERERkDqZPB2bOlI9PnZLLbnz4IZCZCQwcKDtRe3oCixcDDx7I/SZOVKxcMj0MREREVDoJIQPOypXAyZNyW926QFQUMGMGsHs30L49sGFD3ve++KJcxJXoPwxERERU+ggBDBsGrFqVu23iRCAkBBgyBAgLAw4dkrccFSvKoBQfD7z6KjtRkwGuZVZIXMuMiMiEfP890L+/7BAdEgL06JE7qaIQwPnzQOvWwO3bgJUVcPkyUKkSQ5AF4lpmRERUuqWkAJs2Affu5W774w+ga1dgxAj5fPp0OWT+4RmmVSqgVi1g6VIZhl5/HfD2Zhiix+IlMyIiMj3nzwM9ewJ//w3UqCFv587JIfOZmXKfOnVkGCrIq6/K/d3dS6RkKt0YiIiIyPQEB8swBAD//CNvOfr0kSPHmjd/8or0FSsWW4lkXhiIiIjItCQnA7/9Jh///LPsI+TtDbz5JlCmDPDSS7z8RUbHQERERKZl/345sWKVKnKSxcBApSsiC8BO1UREZBp27QLeeAPo3l0+79KFLUFUYthCREREyssZRv+wLl2UqYUsEgMREREpSwhg7lz5uE8fID0dsLYGOnZUti6yKAxERERUsoQA9u6VC6u6ucm5hU6dAhwc5NxBbm5KV0gWiH2IiIioZK1YIVt/AgJk5+nQULl9wACGIVIMW4iIiKjkpKXJ9cQA2Srk4wPExgK2to+fZJGomLGFiIiISs78+cCNG3I+IUCGIQB49105GzWRQthCRERExSc5Wa5C37UrYG8PfPih3P7ll3KW6XXrZAfqnO1ECmEgIiKi4rN4MbBoEbBypew0nZUll90YPFjOMfToUHsihTAQERFR8cjKkqPGALlyfUqKHFm2bBknXCSTw0BERETGtWEDcPOmvCQWHQ1oNPKxrS2wZYtsKSIyMQxERERkPNeuAf36ybmGcrz9tuwjZGsr+xERmSCTHmX24MEDTJ06Fb6+vnBwcMBzzz2Hjz/+GNnZ2fp9hBCYMWMGvLy84ODggDZt2iAqKsrgOOnp6RgzZgzKly8PJycn9OjRA9evXy/p0yEiMn+bN+eGITs7OZT+448BZ2eGITJpJh2I5syZg6VLl2LRokU4d+4c5s6di08//RQLFy7U7zN37lzMmzcPixYtwvHjx6HVatGhQwckJSXp9wkODsbWrVuxfv16HDp0CPfv30dgYCCysrKUOC0iIvN065a8XAYAX3wBJCTIJTkYhKgUUAnxcLumaQkMDISHhwdWrFih39arVy84OjpizZo1EELAy8sLwcHBmDBhAgDZGuTh4YE5c+Zg+PDh0Ol0cHd3x5o1a9C3b18AwM2bN+Ht7Y0dO3agU6dOhaolMTERGo0GOp0OLi4uxj9ZIqLSKDMTuHMHmDVLjijLcf06ULGicnUR/aewf79NuoWoZcuW2LdvH/755x8AwF9//YVDhw6ha9euAIDo6GjExsai40MLAKrVavj7++Pw4cMAgIiICGRmZhrs4+XlBT8/P/0++UlPT0diYqLBjYiIHpKaCjz/PODlZRiGunRhGKJSx6Q7VU+YMAE6nQ61atWCtbU1srKyMGvWLPTr1w8AEPvfDKceHh4G7/Pw8MCVK1f0+9jZ2cHV1TXPPjnvz09oaCg++ugjY54OEZF5mTULOHtWPi5XDvjqK6BBA8DbW9m6iIrApFuINmzYgLVr1+K7777DiRMnEBYWhs8++wxhYWEG+6kemc9CCJFn26OetM+kSZOg0+n0t2vXrhX9RIiIzM3Vq7J/ECA7UsfFAa+9BtSsCTg6KlsbURGYdAvR+++/j4kTJ+L1118HANSrVw9XrlxBaGgogoKCoNVqAchWIE9PT/374uLi9K1GWq0WGRkZiI+PN2gliouLQ/PmzQv8bLVaDbVaXRynRURU+n3/vew/1KoV8MornGiRSj2TbiFKSUmBlZVhidbW1vph976+vtBqtdizZ4/+9YyMDISHh+vDTsOGDWFra2uwT0xMDM6cOfPYQERERI+xcaO8f+MNhiEyCybdQtS9e3fMmjULlStXRt26dXHy5EnMmzcPQ4cOBSAvlQUHByMkJATVq1dH9erVERISAkdHR/T/b30cjUaDYcOGYdy4cShXrhzc3Nwwfvx41KtXD+3bt1fy9IiISqdLl4CICLko6yuvKF0NkVGYdCBauHAhPvzwQ4wcORJxcXHw8vLC8OHDMW3aNP0+H3zwAVJTUzFy5EjEx8ejadOm2L17N5ydnfX7zJ8/HzY2NnjttdeQmpqKgIAArF69GtbW1kqcFhFRycnIAP78E2jaVAYYY8hpHWrTBnB3N84xiRRm0vMQmRLOQ0REpU58PPDyy8DBg8CIEcCSJcY5bsOGwIkTcpHWt982zjGJiolZzENERERFlJEBdOsmwxAgV50/ffrZj/vvvzIM8XIZmRkGIiIiczRxIvC//8mV5mvUkNt69AC2bn264yQlAQsXAn37As2aAdWqye1t2/JyGZkVk+5DREREALKzgSNHZKtP/fqAm9vj9z9+HJg/Xz4OC5OTJbZsCVy5AvTqBezbJwNNfm7cACpUkCvT798vV66/dSvvfv8NXCEyF2whIiIyVVlZcr6f+vWBFi1kiGnUCEhJyX//+Hjg88+BYcPk84EDZR+iKlWA8+dliBFCbk9IyH1fZiZw7x7w3XdylumRI4HERGDAABmGqlUDQkKALVvkJIwbNwJBQcV99kQlii1ERESmJj4eGD8e2LZNBhUAcHGRYSY6GggMBIYOlZfAnJ3lPEB37wLt2wOnTsn9nZyA2bNzj+nkJDtBHz8OXLggg9MnnwCRkUDv3rJvUFaW3Pebb2ToiomRYej0acDBoSS/AaISx1FmhcRRZkRUIm7elK1Bly/L5+XKAcHBwOjRwO7dsi9PjvLlZXCpVAlITpaXuzw8ZGfnV14BHlrUWm/LFnnZrEwZ2Z/olVeA+/cLrmfXLqBTJ2OeIVGJKuzfb7YQERGZki++kGGoShXZUtO6tezPAwB9+shLVgcPAjY2QM4ai//8I++rVgV+/hmoXbvg4/fsKVeoP3UK6NBBbmvVCnj3XSA8HHB1BT7+WG4fM4ZhiCwGW4gKiS1ERFSskpJkp+eAALlQ6pYtjx/WnpoK7N0LeHnJy2ApKcD//V/hLm39+acMV5cvA3XqAH/8AZQtK1+7f19OuOjhIVuQ7OyMcHJEyins328GokJiICKiYiOEbAk6dEg+9/CQrT85LUPFIStLjlzz85ND84nMFC+ZERGVFvv25YYhAHjrreINQ4CcWLFFi+L9DKJShIGIiEgpaWnAV18BK1fK5336yKHx3bopWxeRBWIgIiIqKSkpwNq1wJdfyuHzFSvKIfCA7CT96aeAj4+yNRJZKAYiIqKScOmSnC06JiZ324ULsjPzqFFy0kWGISLFMBARERWnTz+VLUIajQxDlSvLeYXq1QP27JGzSuesNUZEiuEos0LiKDMieqLDh+VSG+++K2d4vntXTp6Yw95ezgyds0AqERU7jjIjIioJSUlA165yVNixY3LG6LAwOY/Q8eO5+zk7A3PmMAwRmSgGIiKiZxEWZjhk3tlZhqS33pKjyADg22/lQqlWXE+byFTxf51EREUlBLBkSe7zPn3kqvIeHnIW6NhYoHp1uf4YwxCRSeP/QomIiiI7G1iwADh7FnB0BBISgB9+ADw9gcmT5T5WVrIFictfEJk8XjIjInpaGzYAU6YA//4rn48YYbj8xfDhcp6hBg2AZs2UqZGIngoDERHRkwgB/PWXXHT1+nXZPyg7G3BxAT75RM4j9DC1Gpg/X5laiahIGIiIiB4nPV0up7Fli+H2wYOBRYsAJydFyiIi42IgIiJ6nBEjZBiytQVq1pR9hZo0AZYtY98gIjPCQEREVJD794HvvpOPt22T8w0RkVniKDMiooLs3QtkZABVqwJduihdDREVIwYiIqKC/PKLvA8MBFQqZWshomLFQERElJ/sbGD7dvk4MFDZWoio2DEQERHlZ9s2OdN02bJA69ZKV0NExYyBiIjoYWfPypFlvXrJ56NHczQZkQXgKDMiohznzwNt28oJGAHAwQF45x1layKiEsFAREQEyPmFunaVYahWLcDVFRg4EHB3V7oyIioBDERERIBcjuPSJaBKFeDAAQYhIgvDPkREZDni4oCmTeVlsEuXgBUrZMvQ338DmzYB1tZyxXqGISKLwxYiIrIcU6cCx47J2+rVQFISMG4cUK2afL1zZ6BxY0VLJCJlsIWIiCzD6dOyRShHUpIcPabTARERclu/fsrURkSKYyAiIsuweLGcbLFHD6BDB6BBAzmqrE6d3H1eflm5+ohIUbxkRkTmLz1d9g0CgOBgoE0b+VilAnbskC1D3bsDZcooVSERKYyBiIjM344dsvN0pUqAv7/humQ+PsDhw4qVRkSmgZfMiMi8paYCs2bJx/37A1b8vz0iyov/z0BE5uvWLbkER0QE4OYGjBypdEVEZKJ4yYyIzNPdu8ALLwAxMXI02dat8vIYEVE+2EJERObpu+9kGKpSRfYR4or1RPQYDEREZJ7WrpX3770HNGyobC1EZPIYiIjI/Fy4IGejtrYG+vZVuhoiKgUYiIjI/KxbJ+87dgQ8PJSthYhKBQYiIjIvQuReLnvjDWVrIaJSg4GIiMzL0aPAv/8CTk5cioOICo2BiIhMw82bQErKsx/n++/l/SuvyFBERFQInIeIiJRx7Bjw229ywsRq1YAuXYCyZYHly5+tZWf3bnnfq5dRyiQiy8BARET5u3MHmDNHzuPTowfg7W28Y9+4AbRrByQny+dWVnIl+rg4oGdPYOZMYPJkwzXHCiM2Fvj7b/k+zjtERE+Bl8yIKH+hocBnnwGjR8tQ9PbbMrQ8i/h4YMIEuchqcjJQowbg4CCP6+QEDBwo95s6VX5eZubTHT88XN43aCBbnoiIComBiIjyt2NH7uPsbODrr4Ft257tmBMmAHPn5j5fuxb4+WegZUtgwwbg22+BxYtli9E338hJFZ/G/v3yvk2bZ6uTiCwOAxGRJYiLA5KSCr9/dLS89GRtLVt1pk6V22fMALKy5ND2mTOBsWOBkyfzP0bO592+Ddy/D1y8CKxcKbe5uwOffgo0bgwEBAAHDwLdusnXRo4ENm6Ujxcvzg05hcFARERFZPKB6MaNG3jjjTdQrlw5ODo64vnnn0dERIT+dSEEZsyYAS8vLzg4OKBNmzaIiooyOEZ6ejrGjBmD8uXLw8nJCT169MD169dL+lSIlBEVBVStKhc2nT4dGDAAuHTp8e/ZuVPeN28uOzq/9x7g4gJERgK1agH9+gEffgjMnw+8+CKwalXuexMSgMBA+b4vvpAdpmvXlpfDsrKArl1lQBs/vuDPf/VVYPhw+bhvX+DMGcPXExJkf6EffgD69wfOnZND7f/5B7CxAfz9n+47IiISJuzevXvCx8dHDB48WBw9elRER0eLvXv3iosXL+r3mT17tnB2dhabN28WkZGRom/fvsLT01MkJibq9xkxYoSoWLGi2LNnjzhx4oRo27ataNCggXjw4EGha9HpdAKA0Ol0Rj1HomKVmipE/fpCyDad3NsLLwiRlpb/e/bvF8LTU+4XEpK7ff16IVxdDY/TvLm8t7ISYvNmITIyhGjSJO/n5dxcXISIiipc7YmJQrz4onyfp6cQ9+/L7ZcuCeHmZnjcypWFmDJFPm7T5pm+MiIyL4X9+23SgWjChAmiZcuWBb6enZ0ttFqtmD17tn5bWlqa0Gg0YunSpUIIIRISEoStra1Yv369fp8bN24IKysrsWvXrkLXwkBEJSY9XYjs7Gc7xp49QkycKERgoAwJ7u5CDBwoRIMGQpQrJ7e1aiXE//5n+L47d4QoU0a+XqOGELduGb6elCTEl18K0bChEB9/LOscNkzub2srxMsv5wakR8OQh4cQR48+3XncvSuEr698/6BBQvTpI0SdOrnHtLXNDW85t88+e5ZvjojMjFkEotq1a4vg4GDRu3dv4e7uLp5//nmxfPly/ev//vuvACBOnDhh8L4ePXqIQYMGCSGE2LdvnwAg7t27Z7BP/fr1xbRp0wpdCwMRFZuLF4VYvlyI774T4swZISpUEMLbW4jhw2UIiIuT+6Wn57aSPC4w3b4tW2JyAoKdnRC//Zb7+vbtMkjkBIqH/mNBTJsmt9evn/tZT/LggQwqD4eSdeuEGDNGPvbxESIzU7YeFcVXX+UNVxqNbGlKThbi/HnDFqO//y7a5xCRWSrs32+Tnofo0qVLWLJkCcaOHYvJkyfj2LFjeOedd6BWqzFo0CDExsYCADweWbzRw8MDV65cAQDExsbCzs4Orq6uefbJeX9+0tPTkZ6ern+emJhorNMiypWdDXToIDsxA4Ctbe5Q82XL5H3ZsrK/TvPmsu/MiBFyfqD582UH5EfNng3k/HutXh34/HOgbdvc17t2lavBv/cesHWr7INTt67sY7Rwodxn6tTCz/JsbS1Hi7VoAfz+O1Cvnuxj1KGD7JA9eLDs11NUgwcD06bJeZEaNZKf06sXUKeOfL1GDTncvnNneb41ahT9s4jIcpVQQCsSW1tb0axZM4NtY8aMES+99JIQQog//vhDABA3b9402OfNN98UnTp1EkIIsW7dOmFnZ5fn2O3btxfDhw8v8LOnT58uAOS5sYWIjOqPP3JbNuztc/vZfPqpvLyV81qLFnlbSRwchDh3ToisrNzjnT4tW4QAIXbufPxnP3iQe0mtTx/5mTmXyp6if12J2LdPiMmT5SW7gmRmPvulRiIyO4VtITLpUWaenp6ok/Nfgf+pXbs2rl69CgDQarUAkKelJy4uTt9qpNVqkZGRgfj4+AL3yc+kSZOg0+n0t2vXrj3z+RABANLT5TDz7Ozc4eVvvAEcOQL07i23jR8vh7PXrStf/+MPQK2Ww9UBwN4eSE2Vo7dq1ZLLYIwcKZe8yMiQo7w6dXp8HdbWcvJFQH7m++/LxxMmyNdMSbt2wKxZQJkyBe9jY/P0M1sTEf2nSIFo6NChSMpnTpPk5GQMHTr0mYvK0aJFC5w/f95g2z///AMfHx8AgK+vL7RaLfbs2aN/PSMjA+Hh4WjevDkAoGHDhrC1tTXYJyYmBmfOnNHvkx+1Wg0XFxeDG5FRjBwpl5UYMyY3EPXpI2dX3rgR6NhRblOpgI8+ko9r15aXtw4flmt9/e9/uTMxX7gANG0KLFkiL71ptXK+n8KEAz8/OQw/R6VKMpwREVmaojQ/WVlZiVuPjj4RQty+fVtYW1sX5ZD5OnbsmLCxsRGzZs0SFy5cEOvWrROOjo5i7dq1+n1mz54tNBqN2LJli4iMjBT9+vXLd9h9pUqVxN69e8WJEydEu3btOOyenk1WlhAHDshh7YX188+5nZYfvrm6Pv449+4ZXhbLkZkpxKFDQlhby+PUri3E7NmFH9aeIzVViM8/F+Kll2SNRERmpFhGmel0OpGQkCBUKpW4ePGi0Ol0+tu9e/dEWFiY8PT0fKbCH/Xzzz8LPz8/oVarRa1atQxGmQkhh95Pnz5daLVaoVarRevWrUVkZKTBPqmpqWL06NHCzc1NODg4iMDAQHH16tWnqoOByMIlJOSGluxsIYYOlSHkrbdy9/nnHyEGD5b9eB7100955+PJGS21b9+z1bZ2rRB9+wpx7dqzHYeIyAwV9u+3SgghCtuaZGVlBdVjmuFVKhU++ugjTJky5ZlbrkxNYmIiNBoNdDodL59Zmn37gPbt5SWoBg1kP56HLsHi+nWgXDm5DMWZM4Cvr+z/o9EAaWly+YnQUODuXcDREbCzk32IIiPlpa7nnlPu3IiIzFxh/34/1VjY/fv3QwiBdu3aYfPmzXB7aDVpOzs7+Pj4wMvLq+hVE5miDRvkvRDAqVN5X2/fXgafy5fl8+hoIDhYLmcxfLhcsBQAXnhBdo62spIdpP38SqB4IiIqjKdqIcpx5coVVK5c+bGtReaGLUQW5PJlYPJk4MQJwMsLOH8euHlTdma2tgZ0OhmCrlwBunfPfZ9aDUycKDtCq1Rybp6cDsvz5gHDhsn1wIiIqMQU9u93kQIRABw8eBDLli3DpUuXsHHjRlSsWBFr1qyBr68vWrZsWeTCTRUDkRk5d05OPLhvn2y96dFDbo+Olqu5b9woh7Q/TKWSl7wemeATmzbJSRB9fOQQea1WDn3/6afcfQYMkOGIiIhKXGH/fhdp2P3mzZvRqVMnODg44MSJE/oZnZOSkhASElK0iolKws6dQJMmwJYtsqXnnXfkvECxsXI252+/lWGoVSu54noOP7+8YQiQ8wYNHQoEBMgwBMhZpXP4+OTO9UNERCarSIFo5syZWLp0Kb7++mvY2trqtzdv3hwnTpwwWnFERqXTyQBz/77szAzIy16zZwPdusnH1aoBhw7JpSBy+v4AsqN0YTVqBKxeLS+TRUUB3t5GPQ0iIjK+IgWi8+fPo3Xr1nm2u7i4ICEh4VlrIioe+/YBKSky9Bw4AHz1ldw+Y4bsL1S+vGxBatFCXiJzcgIWLZIjyKZNe7rPCgqSa4UVdj0wIiJSVJECkaenJy5evJhn+6FDh/AchxCTqfr1V3nftasc+j58OPDBB3KbgwPwyy8yLD1s1Ci5qGjDhiVbKxERlagiBaLhw4fj3XffxdGjR6FSqXDz5k2sW7cO48ePx8j8Vt8mUpoQuYEoZ40vKyu5avypU8Bff+VeRiMiIovzVPMQ5fjggw+g0+nQtm1bpKWloXXr1lCr1Rg/fjxGjx5t7BqJnk12tpwc8coV2TLk72/4eoMGytRFREQm46mH3WdlZeHQoUOoV68e7O3tcfbsWWRnZ6NOnToo87iVqEs5DrsvxUJD5bxCANC/P7BunbL1EBFRiSnWeYjs7e1x7tw5+D7NyJtSjoGolMjKkpe/6tcHbGyAzEw59D0mRoaiDz8E7O2VrpKIiEpIsc5DVK9ePVy6dKnIxREVi3v3gC5dZAfoTp3kzc5OhqEKFYDp0xmGiIgoX0XqQzRr1iyMHz8en3zyCRo2bAinR4YWswWFSty9e3JixdOn5fPffjN8fcgQGY6IiIjyUaRLZlZWuQ1LD69nJoSASqVCVlaWcaozIbxkZqLu3JGjw3JaLLVauZbYxIlAzZpyksR//pGTLHp4KFsrERGVuGJZ7T7H/v37i1wYkVGtXJkbhipWBHbtkstsDBkCPDSLOhER0eMUKRD5PzpsmaikCCHXDjt5Eli2DPj6a7k9JAQYPz43BDEMERHRUyhSIDqd00/jESqVCvb29qhcuTLUavUzFUaUr0OH5DphAPDSS/Le2Vku0soQRERERVSkQPT8888b9B16lK2tLfr27Ytly5bBnqN66FmkpQE3bshRYs7O+a8cP3w41wwjIqJnUqRh91u3bkX16tWxfPlynDp1CidPnsTy5ctRs2ZNfPfdd1ixYgV+++03TJ061dj1kiVJSgKaNZPri7m6AoGBcvFVKyvg4kUZlM6dk6vVExERPYMiD7v/4osv0ClnTSgA9evXR6VKlfDhhx/i2LFjcHJywrhx4/DZZ58ZrViyIEIAAwfKdcZUKjnh4vbt8rWpU4GqVeVjLy/FSiQiIvNRpEAUGRkJHx+fPNt9fHwQGRkJQF5Wi4mJebbqyHJ9/TXw44+AWg2Eh8vJFWfOBAYMAN57T+nqiIjIzBQpENWqVQuzZ8/G8uXLYfffZHeZmZmYPXs2atWqBQC4ceMGPDjvCxXF9etyxBgg+wzlrELfs6diJRERkXkrUiBavHgxevTogUqVKqF+/fpQqVQ4ffo0srKy8MsvvwAALl26hJEjRxq1WDIzO3YAYWFyfTE/v9ztEybI/kNNm8rRY0RERMWsSDNVA8D9+/exdu1a/PPPPxBCoFatWujfvz+cnZ2NXaNJ4EzVRvbJJ8C0afJx48bA0aOyr9ChQ0CrVvLxn38CL76obJ1ERFSqFetq95aIgciIdu4EunY13LZ5s5xXqEkTOXps2DDgm2+UqY+IiMxGsa52DwBr1qxBy5Yt4eXlhStXrgAA5s+fjx9//LGohyRL8McfsmM0AIwZI0eMAcDbbwP+/jIM1aoFcHQiERGVoCIFoiVLlmDs2LHo0qUL4uPj9Yu5urq6YsGCBcasj8xJVBTQvj0QHy/nF5o7F/jgA3nJ7O5dObdQpUrAL78AZcsqXS0REVmQIgWihQsX4uuvv8aUKVNgY5PbL7tRo0b6YfdEeunpwN9/A++/L2eebtcO2LsXsLeXs0/v2wcEBckFWU+dyp1jiIiIqIQUaZRZdHQ0XnjhhTzb1Wo1kpOTn7koMiNCAK++KkeUAYCNDbB0KeDomLuPs3Pu+mREREQKKFILka+vL06dOpVn+86dO1G7du1nrYnMyfff54YhQPYbql5duXqIiIjyUaQWovfffx+jRo1CWloahBA4duwYvv/+e4SEhGDFihXGrpFKq8WL5ZxCgOw83asXUK+esjURERHlo0iBaMiQIXjw4AE++OADpKSkoH///qhYsSIWLlyIVq1aGbtGKo22bQNGj5aPAwJkIFKrFS2JiIioIEUedv/WW2/hypUriIuLQ2xsLI4dO4aTJ0+iWrVqxqyPSqOMjNylN955B9i9m2GIiIhM2lMFooSEBAwYMADu7u7w8vLCl19+CTc3NyxevBjVqlXDkSNHsHLlyuKqlUqLRYuAf/8FtFpg1izAqsi5m4iIqEQ81SWzyZMn48CBAwgKCsKuXbvw3nvvYdeuXUhLS8OOHTvg7+9fXHVSaXHnDvDxx/LxzJlAmTLK1kNERFQITxWItm/fjlWrVqF9+/YYOXIkqlWrhho1anAyRsr10UeATgc8/zwweLDS1RARERXKU13LuHnzJurUqQMAeO6552Bvb48333yzWAqjUujcOWDJEvn4888Ba2tl6yEiIiqkpwpE2dnZsLW11T+3traGk5OT0YuiUur994GsLKBHDzkbNRERUSnxVJfMhBAYPHgw1P+NGEpLS8OIESPyhKItW7YYr0IqHf78E9i+XbYKffqp0tUQERE9lacKREFBQQbP33jjDaMWQ6WQEMCRI8CMGfJ5//5AjRqKlkRERPS0VEIIoXQRpUFiYiI0Gg10Oh1cXFyULsd0rFoFDB0qH6tUckV7Lt9CREQmorB/vzlBDBXe2bNAx47AwoWyZejBA+CTT+Rr1avLYfYMQ0REVAoVaekOshBCyOU3VCpg1CigUyfg+nVgzx5g/37AyQmIjgbc3YFTpwxXsCciIipFGIioYMeOAV99JR8vXizvPT2BmBhg69bc/T74gGGIiIhKNQYiKtjDoQcAWrYEvvtOthLt2iUnYGzZUq5iT0REVIoxEFHBtm2T96Ghcubpjh3lumTe3kCzZkpWRkREZFQMRJS/X38Fzp8H7OyAkSMBjqwjIiIzxlFmlNdPPwFdu8rHPXsyDBERkdljIKK8Fi0CsrOB3r2BFSuUroaIiKjYMRCRofv3gfBw+fiTT4AyZZSth4iIqAQwEJGh334DMjIAX1+gZk2lqyEiIioRDERkaMcOed+1q5yQkYiIyAIwEFGurCzgxx/l45xO1URERBaAgYhy/f47EBsLuLkB7dsrXQ0REVGJKVWBKDQ0FCqVCsHBwfptQgjMmDEDXl5ecHBwQJs2bRAVFWXwvvT0dIwZMwbly5eHk5MTevTogevXr5dw9aXAd9/J+z595PxDREREFqLUBKLjx49j+fLlqF+/vsH2uXPnYt68eVi0aBGOHz8OrVaLDh06ICkpSb9PcHAwtm7divXr1+PQoUO4f/8+AgMDkZWVVdKnYbrS0oDNm+Xj/v2VrYWIiKiElYpAdP/+fQwYMABff/01XF1d9duFEFiwYAGmTJmCV199FX5+fggLC0NKSgq++6+1Q6fTYcWKFfj888/Rvn17vPDCC1i7di0iIyOxd+9epU7J9OzcKdcmq1RJrk9GRERkQUpFIBo1ahS6deuG9o/0a4mOjkZsbCw6duyo36ZWq+Hv74/Dhw8DACIiIpCZmWmwj5eXF/z8/PT75Cc9PR2JiYkGN7OWc7msXz+5XhkREZEFMfm/fOvXr0dERARCQ0PzvBYbGwsA8PDwMNju4eGhfy02NhZ2dnYGLUuP7pOf0NBQaDQa/c3b2/tZT8U0HT0KNGgAbNokn/NyGRERWSCTDkTXrl3Du+++i3Xr1sHe3r7A/VSPzJcjhMiz7VFP2mfSpEnQ6XT627Vr156u+NIgPl52oD59Wj7385PhiIiIyMKYdCCKiIhAXFwcGjZsCBsbG9jY2CA8PBxffvklbGxs9C1Dj7b0xMXF6V/TarXIyMhAfHx8gfvkR61Ww8XFxeBmdsaOBa5dA6pWBb75Rs5BxMkYiYjIApl0IAoICEBkZCROnTqlvzVq1AgDBgzAqVOn8Nxzz0Gr1WLPnj3692RkZCA8PBzNmzcHADRs2BC2trYG+8TExODMmTP6fSzS2bNAWJh8vGYNMGwY8NxzytZERESkEBulC3gcZ2dn+Pn5GWxzcnJCuXLl9NuDg4MREhKC6tWro3r16ggJCYGjoyP6/9cXRqPRYNiwYRg3bhzKlSsHNzc3jB8/HvXq1cvTSduiTJ8OCAG8+irQrJnS1RARESnKpANRYXzwwQdITU3FyJEjER8fj6ZNm2L37t1wdnbW7zN//nzY2NjgtddeQ2pqKgICArB69WpYW1srWLmCEhKALVvk448+UrQUIiIiU6ASQgiliygNEhMTodFooNPpSn9/ol9+Abp3B2rUAM6fV7oaIiKiYlPYv98m3YeIikl4uLxv3VrZOoiIiEwEA5ElyglE/v7K1kFERGQiGIgsTVIScOKEfMxAREREBICByPJs3w5kZckh9uY6+zYREdFTYiCyNEuXyvtBg5Stg4iIyIQwEFmSw4dl/yFra+DNN5WuhoiIyGQwEFmKL74AWraUj19+GahYUdl6iIiITAgDkSW4cQOYNEnOTP3KK8BXXyldERERkUkp9TNVUyHMmAGkpgItWgCbN3MBVyIiokewhcjcnTsHrFwpH8+dyzBERESUDwYiczdpEpCdDfTsCTRvrnQ1REREJomByJydOQP8+CNgZQWEhChdDRERkcliIDJnmzfL+65dgdq1la2FiIjIhDEQmbMtW+R9r17K1kFERGTiGIjM1b//AqdPy0kYu3dXuhoiIiKTxkBkrnKW6PD3B8qVU7YWIiIiE8dAZI4uXpQzUwPA2LHK1kJERFQKMBCZow8/BDIzgY4dZYdqIiIieiwGInPz77/ADz/Ix3PmcCJGIiKiQmAgMgcpKfJepwMmTpQTMXbpAjz/vKJlERERlRYMRKXdvHmAs7Ncr6xWLWDTJrl94kRFyyIiIipNVEIIoXQRpUFiYiI0Gg10Oh1cXFyULkcSAlCrZX+hHFWrAqGhQJ8+ytVFRERkIgr795ur3ZdmUVGGYcjNDfjtN6ByZeVqIiIiKoUYiEqzrVvlfatWckRZ9+4MQ0REREXAQFRaJSQAq1bJx0FBwLBhipZDRERUmrFTdWmUlibXJ4uOBipW5FplREREz4iBqLRJSwNefln2FXJyAn75BShbVumqiIiISjUGotJm+nRg924ZhrZv51xDRERERsBAVJocPw589pl8vG6dXLiViIiInhkDUWkyebKchbpfP3nZjIiIiIyCgai0OHkS2LsXsLYGQkKUroaIiMisMBCVBjdvAu++Kx+/9hpQpYqi5RAREZkbzkNk6v79F2jWDLh9Wy7TMWmS0hURERGZHbYQmbI7d4Bu3WQY8vOTnarr1VO6KiIiIrPDFiJTFRcHtG0LnD8PeHvLofaenkpXRUREZJbYQmSqpk4Fzp6VM1Hv2cMwREREVIwYiEyRTifnGQLkfc2aytZDRERk5hiITNG33wIpKbLfUOvWSldDRERk9hiITNGaNfJ+xAhApVK2FiIiIgvAQGRqdDogIkI+5mzUREREJYKByNT88YdcnqNqVaBSJaWrISIisggMRKYmPFzec+FWIiKiEsNAZGp+/13eMxARERGVGAYiU5Kdndt/qGVLZWshIiKyIAxEpiQ9HcjKko/d3ZWthYiIyIIwEJmStLTcx/b2ytVBRERkYRiITElOILKyAmy4zBwREVFJYSAyJTmByN6eEzISERGVIAYiU/JwICIiIqISw0BkShiIiIiIFMFAZErS0+U9AxEREVGJYiAyJWwhIiIiUgQDkSlhICIiIlKESQei0NBQNG7cGM7OzqhQoQJ69uyJ8+fPG+wjhMCMGTPg5eUFBwcHtGnTBlFRUQb7pKenY8yYMShfvjycnJzQo0cPXL9+vSRPpXAYiIiIiBRh0oEoPDwco0aNwpEjR7Bnzx48ePAAHTt2RHJysn6fuXPnYt68eVi0aBGOHz8OrVaLDh06ICkpSb9PcHAwtm7divXr1+PQoUO4f/8+AgMDkZUzK7SpYCAiIiJShEoIIZQuorBu376NChUqIDw8HK1bt4YQAl5eXggODsaECRMAyNYgDw8PzJkzB8OHD4dOp4O7uzvWrFmDvn37AgBu3rwJb29v7NixA506dSrUZycmJkKj0UCn08HFxaV4TnD1amDIEKBrV2D79uL5DCIiIgtS2L/fJt1C9CidTgcAcHNzAwBER0cjNjYWHTt21O+jVqvh7++Pw4cPAwAiIiKQmZlpsI+Xlxf8/Pz0++QnPT0diYmJBrdixxYiIiIiRZSaQCSEwNixY9GyZUv4+fkBAGJjYwEAHh4eBvt6eHjoX4uNjYWdnR1cXV0L3Cc/oaGh0Gg0+pu3t7cxTyd/DERERESKKDWBaPTo0Th9+jS+//77PK+pHlnmQgiRZ9ujnrTPpEmToNPp9Ldr164VrfCnwUBERESkiFIRiMaMGYOffvoJ+/fvR6VKlfTbtVotAORp6YmLi9O3Gmm1WmRkZCA+Pr7AffKjVqvh4uJicCt2DERERESKMOlAJITA6NGjsWXLFvz222/w9fU1eN3X1xdarRZ79uzRb8vIyEB4eDiaN28OAGjYsCFsbW0N9omJicGZM2f0+5gMBiIiIiJF2ChdwOOMGjUK3333HX788Uc4OzvrW4I0Gg0cHBygUqkQHByMkJAQVK9eHdWrV0dISAgcHR3Rv39//b7Dhg3DuHHjUK5cObi5uWH8+PGoV68e2rdvr+Tp5cVAREREpAiTDkRLliwBALRp08Zg+6pVqzB48GAAwAcffIDU1FSMHDkS8fHxaNq0KXbv3g1nZ2f9/vPnz4eNjQ1ee+01pKamIiAgAKtXr4a1tXVJnUrhMBAREREpolTNQ6SkEpmHaPBgICwMmDsXeP/94vkMIiIiC2KW8xCZPbYQERERKYKByJQwEBERESmCgciUMBAREREpgoHIlDAQERERKYKByJQwEBERESmCgciUMBAREREpgoHIlDAQERERKYKByJQwEBERESmCgciUMBAREREpgoHIlDAQERERKYKByJQwEBERESmCgchUCAGkp8vHDEREREQlioHIVOSEIYCBiIiIqIQxEJmKnMtlAAMRERFRCWMgMhU5gcjKCrCxUbYWIiIiC8NAZCoe7lCtUilbCxERkYVhIDIVHGFGRESkGAYiU8FAREREpBgGIlPBQERERKQYBiJTkZoq7xmIiIiIShwDkalITpb3ZcooWwcREZEFYiAyFffvy3snJ2XrICIiskAMRKYip4WIgYiIiKjEMRCZCgYiIiIixTAQmQoGIiIiIsUwEJkKBiIiIiLFMBCZCgYiIiIixTAQmQoGIiIiIsUwEJkKBiIiIiLFMBCZCgYiIiIixTAQmQoGIiIiIsUwEJkKzlRNRESkGAYiU8EWIiIiIsUwEJkKLu5KRESkGAYiU8EWIiIiIsUwEJkKBiIiIiLFMBCZAiEYiIiIiBTEQGQK0tJkKAIYiIiIiBTAQGQKclqHAMDRUbk6iIiILBQDkSnICUT29oC1tbK1EBERWSAGIlPA/kNERESKYiAyBZylmoiISFEMRKaALURERESKYiAyBQxEREREimIgMgUMRERERIpiIDIFDERERESKYiAyBVzYlYiISFEMRKaAo8yIiIgUxUBkCi5elPeVKilbBxERkYViIDIFkZHyvl49ZesgIiKyUAxESsvOBqKi5GMGIiIiIkUwECktOhpISQHUaqBaNaWrISIiskgMRErLuVxWpw5gY6NsLURERBbKogLRV199BV9fX9jb26Nhw4Y4ePCg0iWx/xAREZEJsJhAtGHDBgQHB2PKlCk4efIkWrVqhS5duuDq1avKFsZAREREpDiLCUTz5s3DsGHD8Oabb6J27dpYsGABvL29sWTJEmULS00FrK0ZiIiIiBRkEYEoIyMDERER6Nixo8H2jh074vDhwwpV9Z+ff5YTM7Ztq2wdREREFswievHeuXMHWVlZ8PDwMNju4eGB2NjYfN+Tnp6O9PR0/fPExMTiK9DevviOTURERE9kES1EOVQqlcFzIUSebTlCQ0Oh0Wj0N29v75IokYiIiBRgEYGofPnysLa2ztMaFBcXl6fVKMekSZOg0+n0t2vXrpVEqURERKQAiwhEdnZ2aNiwIfbs2WOwfc+ePWjevHm+71Gr1XBxcTG4ERERkXmyiD5EADB27FgMHDgQjRo1QrNmzbB8+XJcvXoVI0aMULo0IiIiUpjFBKK+ffvi7t27+PjjjxETEwM/Pz/s2LEDPj4+SpdGREREClMJIYTSRZQGiYmJ0Gg00Ol0vHxGRERUShT277dF9CEiIiIiehwGIiIiIrJ4DERERERk8RiIiIiIyOIxEBEREZHFYyAiIiIii8dARERERBbPYiZmfFY50zUV66r3REREZFQ5f7efNO0iA1EhJSUlAQBXvSciIiqFkpKSoNFoCnydM1UXUnZ2Nm7evAlnZ2eoVCqjHTcxMRHe3t64du2aRc2AzfPmeVsCnjfP2xKY+nkLIZCUlAQvLy9YWRXcU4gtRIVkZWWFSpUqFdvxXVxcTPIfUnHjeVsWnrdl4XlbFlM+78e1DOVgp2oiIiKyeAxEREREZPEYiBSmVqsxffp0qNVqpUspUTxvnrcl4HnzvC2BuZw3O1UTERGRxWMLEREREVk8BiIiIiKyeAxEREREZPEYiIiIiMjiMRAp7KuvvoKvry/s7e3RsGFDHDx4UOmSjGbGjBlQqVQGN61Wq39dCIEZM2bAy8sLDg4OaNOmDaKiohSsuGgOHDiA7t27w8vLCyqVCtu2bTN4vTDnmZ6ejjFjxqB8+fJwcnJCjx49cP369RI8i6f3pPMePHhwnt//pZdeMtinNJ53aGgoGjduDGdnZ1SoUAE9e/bE+fPnDfYxx9+8MOdtjr/5kiVLUL9+ff2kg82aNcPOnTv1r5vjbw08+bzN8bdmIFLQhg0bEBwcjClTpuDkyZNo1aoVunTpgqtXrypdmtHUrVsXMTEx+ltkZKT+tblz52LevHlYtGgRjh8/Dq1Wiw4dOujXjSstkpOT0aBBAyxatCjf1wtznsHBwdi6dSvWr1+PQ4cO4f79+wgMDERWVlZJncZTe9J5A0Dnzp0Nfv8dO3YYvF4azzs8PByjRo3CkSNHsGfPHjx48AAdO3ZEcnKyfh9z/M0Lc96A+f3mlSpVwuzZs/Hnn3/izz//RLt27fDyyy/rQ485/tbAk88bML/fGoIU06RJEzFixAiDbbVq1RITJ05UqCLjmj59umjQoEG+r2VnZwutVitmz56t35aWliY0Go1YunRpCVVofADE1q1b9c8Lc54JCQnC1tZWrF+/Xr/PjRs3hJWVldi1a1eJ1f4sHj1vIYQICgoSL7/8coHvMYfzFkKIuLg4AUCEh4cLISznN3/0vIWwnN/c1dVVfPPNNxbzW+fIOW8hzPO3ZguRQjIyMhAREYGOHTsabO/YsSMOHz6sUFXGd+HCBXh5ecHX1xevv/46Ll26BACIjo5GbGyswfmr1Wr4+/ub1fkX5jwjIiKQmZlpsI+Xlxf8/PxK/Xfx+++/o0KFCqhRowbeeustxMXF6V8zl/PW6XQAADc3NwCW85s/et45zPk3z8rKwvr165GcnIxmzZpZzG/96HnnMLffmou7KuTOnTvIysqCh4eHwXYPDw/ExsYqVJVxNW3aFN9++y1q1KiBW7duYebMmWjevDmioqL055jf+V+5ckWJcotFYc4zNjYWdnZ2cHV1zbNPaf630KVLF/Tp0wc+Pj6Ijo7Ghx9+iHbt2iEiIgJqtdoszlsIgbFjx6Jly5bw8/MDYBm/eX7nDZjvbx4ZGYlmzZohLS0NZcqUwdatW1GnTh39H3Zz/a0LOm/APH9rBiKFqVQqg+dCiDzbSqsuXbroH9erVw/NmjVD1apVERYWpu98Z87n/7CinGdp/y769u2rf+zn54dGjRrBx8cH27dvx6uvvlrg+0rTeY8ePRqnT5/GoUOH8rxmzr95Qedtrr95zZo1cerUKSQkJGDz5s0ICgpCeHi4/nVz/a0LOu86deqY5W/NS2YKKV++PKytrfMk5bi4uDz/tWEunJycUK9ePVy4cEE/2szcz78w56nVapGRkYH4+PgC9zEHnp6e8PHxwYULFwCU/vMeM2YMfvrpJ+zfvx+VKlXSbzf337yg886PufzmdnZ2qFatGho1aoTQ0FA0aNAAX3zxhdn/1gWdd37M4bdmIFKInZ0dGjZsiD179hhs37NnD5o3b65QVcUrPT0d586dg6enJ3x9faHVag3OPyMjA+Hh4WZ1/oU5z4YNG8LW1tZgn5iYGJw5c8asvou7d+/i2rVr8PT0BFB6z1sIgdGjR2PLli347bff4Ovra/C6uf7mTzrv/JjLb/4oIQTS09PN9rcuSM5558csfusS78ZNeuvXrxe2trZixYoV4uzZsyI4OFg4OTmJy5cvK12aUYwbN078/vvv4tKlS+LIkSMiMDBQODs7689v9uzZQqPRiC1btojIyEjRr18/4enpKRITExWu/OkkJSWJkydPipMnTwoAYt68eeLkyZPiypUrQojCneeIESNEpUqVxN69e8WJEydEu3btRIMGDcSDBw+UOq0netx5JyUliXHjxonDhw+L6OhosX//ftGsWTNRsWLFUn/e//d//yc0Go34/fffRUxMjP6WkpKi38ccf/Mnnbe5/uaTJk0SBw4cENHR0eL06dNi8uTJwsrKSuzevVsIYZ6/tRCPP29z/a0ZiBS2ePFi4ePjI+zs7MSLL75oMIS1tOvbt6/w9PQUtra2wsvLS7z66qsiKipK/3p2draYPn260Gq1Qq1Wi9atW4vIyEgFKy6a/fv3CwB5bkFBQUKIwp1namqqGD16tHBzcxMODg4iMDBQXL16VYGzKbzHnXdKSoro2LGjcHd3F7a2tqJy5coiKCgozzmVxvPO75wBiFWrVun3Mcff/Ennba6/+dChQ/X/H+3u7i4CAgL0YUgI8/ythXj8eZvrb60SQoiSa48iIiIiMj3sQ0REREQWj4GIiIiILB4DEREREVk8BiIiIiKyeAxEREREZPEYiIiIiMjiMRARERGRxWMgIiIiIovHQEREFq1NmzYIDg5WugwiUhgDEREREVk8Lt1BRBZr8ODBCAsLM9gWHR2NKlWqKFMQESmGgYiILJZOp0OXLl3g5+eHjz/+GADg7u4Oa2trhSsjopJmo3QBRERK0Wg0sLOzg6OjI7RardLlEJGC2IeIiIiILB4DEREREVk8BiIismh2dnbIyspSugwiUhgDERFZtCpVquDo0aO4fPky7ty5g+zsbKVLIiIFMBARkUUbP348rK2tUadOHbi7u+Pq1atKl0RECuCweyIiIrJ4bCEiIiIii8dARERERBaPgYiIiIgsHgMRERERWTwGIiIiIrJ4DERERERk8RiIiIiIyOIxEBEREZHFYyAiIiIii8dARERERBaPgYiIiIgsHgMRERERWbz/B1irGk70QxWcAAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(0)\n",
    "plt.xlabel(\"t\")\n",
    "plt.ylabel(\"Regret\")\n",
    "plt.title(\"All classes with 3 Learners\")\n",
    "plt.plot(np.cumsum(opt123-ts_all_classes_rewards, axis=0), 'r')\n",
    "plt.legend([\"TS\"])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 303,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_______________________________________________\n",
      "Considering all the experiments:\n",
      "-----------------------------------------------\n",
      "Total expected reward over 365 days\n",
      "44817.87232016667\n",
      "-----------------------------------------------\n",
      "Total expected regret over 365 days\n",
      "1050.9966498892672\n",
      "-----------------------------------------------\n",
      "Average expected reward per round over 365 days\n",
      "122.78869128812786\n",
      "-----------------------------------------------\n",
      "Average expected regret per round over 365 days\n",
      "2.8794428764089512\n"
     ]
    }
   ],
   "source": [
    "print(\"_______________________________________________\")\n",
    "print(\"Considering all the experiments:\")\n",
    "print(\"-----------------------------------------------\")\n",
    "print(\"Total expected reward over\",n_days, \"days\")\n",
    "print(np.mean(np.sum(ts_all_classes_rewards)))\n",
    "print(\"-----------------------------------------------\")\n",
    "print(\"Total expected regret over\",n_days, \"days\")\n",
    "print(np.mean(np.sum(opt123 - ts_all_classes_rewards)))\n",
    "\n",
    "print(\"-----------------------------------------------\")\n",
    "print(\"Average expected reward per round over\",n_days, \"days\")\n",
    "print(np.mean(np.mean(ts_all_classes_rewards, axis=0)))\n",
    "print(\"-----------------------------------------------\")\n",
    "print(\"Average expected regret per round over\",n_days, \"days\")\n",
    "print(np.mean(np.mean(opt123-ts_all_classes_rewards, axis=0)))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "TS second split, 3 classes (1 leaner, all class at the same time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 312,
   "metadata": {},
   "outputs": [],
   "source": [
    "#3 classes together (simultaneously)\n",
    "\n",
    "E = E_commerce(binary_features=1)\n",
    "E.set_lambda(0.5)\n",
    "E.set_products(products)\n",
    "E.graph = np.array([[0. , 0.5, 0. , 1. , 0. ],[1. , 0. , 0. , 0.5, 0. ],[0. , 1. , 0. , 0.5, 0. ],[0. , 0.5, 1. , 0. , 0. ],[1. , 0. , 0. , 0.5, 0. ]])\n",
    "\n",
    "n_arms = 4\n",
    "num_users = 100\n",
    "n_days = 365\n",
    "fixed_alpha = 0\n",
    "fixed_weights = 1\n",
    "fixed_units = 0\n",
    "n_experiments = 10\n",
    "\n",
    "ts_rewards_per_experiment = []\n",
    "\n",
    "\n",
    "ts_pulls_per_arm_per_experiment = []\n",
    "\n",
    "\n",
    "for e in range(n_experiments):\n",
    "\n",
    "    env = Environment(n_arms, E, margins_matrix, num_users, fixed_alpha,fixed_weights,fixed_units)\n",
    "    ts_learner = TS_poisson(n_arms=n_arms)\n",
    "\n",
    "    for d in range(n_days):\n",
    "        \n",
    "        pulled_arm = ts_learner.pull_arm(env.margins_matrix)\n",
    "        reward = env.round(pulled_arm)\n",
    "        ts_learner.update(pulled_arm, reward, env.clicks_current_day, env.purchases_current_day, env.daily_units)\n",
    "\n",
    "    ts_rewards_per_experiment.append(ts_learner.collected_rewards)\n",
    "\n",
    "    ts_pulls_per_arm_per_experiment.append(ts_learner.counter_per_arm)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 313,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "118.17314762734482"
      ]
     },
     "execution_count": 313,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#we don't use opt123 but opt_3_Classes because now they are all the 3 classes at the same time (we use the opt estimated by the Conv Rates learner which considered all the classes at the same time)\n",
    "opt_3_classes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 314,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAkQAAAHFCAYAAAAT5Oa6AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/MnkTPAAAACXBIWXMAAA9hAAAPYQGoP6dpAABUFklEQVR4nO3deVhUVR8H8O/IMiyyCAoDioiKK265JZqguKSiqZl7Ylq5F6mvqb2mluFSqeWauaYZLWpmLrkRSi4puWKZvqKigLjggLLDef84MTYCCgjcYeb7eZ55Zu6dMzO/O+MT38499xyVEEKAiIiIyIRVULoAIiIiIqUxEBEREZHJYyAiIiIik8dARERERCaPgYiIiIhMHgMRERERmTwGIiIiIjJ5DERERERk8hiIiIiIyOQxEBGVgM8//xwqlQo+Pj4FtlGpVJg1a5Zu+9dff4VKpcKvv/76zJ9/9epVqFQqrF+//pnfy9j4+/vD399ft52SkoJZs2bl+73PmjULKpUKd+7cKbsCS5GxHQ9RaTJXugAiY7B27VoAQFRUFI4fP47WrVsrXBHlWr58ud52SkoKZs+eDQB6QYmITBt7iIie0cmTJ3HmzBn06NEDALBmzRqFK6J/a9CgARo0aKB0GSYtOzsb6enpZfZ5KSkpZfZZZDwYiIieUW4AmjdvHnx9fREaGlri/0G+efMm3nzzTXh4eMDS0hLu7u7o168fbt26VeBrLl++jNdeew3e3t6wsbFB1apV0bNnT5w7d06vXU5ODubMmYO6devC2toajo6OaNy4MT777DNdm9u3b+s+X61Wo0qVKmjbti3279+v91779+9HQEAA7O3tYWNjg7Zt2+LAgQN6bQr7Xv8WFRUFlUqF77//XrcvMjISKpUKDRs21Gvbq1cvNG/eXLf971NmV69eRZUqVQAAs2fPhkqlgkqlwvDhw/Xe49atWxg0aBAcHBzg6uqKESNGQKvVFljfv61duxZNmjSBlZUVnJyc0KdPH/z55596bYYPH46KFSvi8uXL6N69OypWrAgPDw9MmjQpT3DIyMjAnDlzUK9ePd339dprr+H27duFqqcwTp48iV69esHJyQlWVlZo1qwZvvvuO702t2/fxtixY9GgQQNUrFgRLi4u6NixIw4fPqzXLvf07YIFCzBnzhx4eXlBrVYjLCxMdwovKirqqd+vEALLly9H06ZNYW1tjUqVKqFfv364cuWKXjt/f3/4+Pjg0KFD8PX1hY2NDUaMGFFi3w2ZDgYiomeQmpqKb775Bi1btoSPjw9GjBiB5ORkvT/cz+rmzZto2bIltm3bhokTJ2L37t1YvHgxHBwckJiYWODrYmNj4ezsjHnz5mHPnj1YtmwZzM3N0bp1a1y8eFHXbsGCBZg1axYGDRqEnTt34ttvv8XIkSNx//59XZtXX30VP/74I95//33s3bsXq1evRqdOnXD37l1dm02bNqFLly6wt7fHhg0b8N1338HJyQldu3bVC0WFea/HNWzYEG5ubnqhaf/+/bC2tsaFCxcQGxsLAMjKykJ4eDg6deqU7/u4ublhz549AICRI0fi6NGjOHr0KGbMmKHX7uWXX0adOnWwZcsWTJ06FZs3b8Y777xTYH255s6di5EjR6Jhw4bYunUrPvvsM5w9exZt2rTBpUuX9NpmZmaiV69eCAgIwPbt2zFixAgsWrQI8+fP17XJycnBSy+9hHnz5mHw4MHYuXMn5s2bh3379sHf3x+pqalPrelpwsLC0LZtW9y/fx8rV67E9u3b0bRpUwwYMEBvTNq9e/cAADNnzsTOnTuxbt061KxZE/7+/vmOx/r8889x8OBBfPLJJ9i9ezfq1aune64w3++oUaMQHByMTp064ccff8Ty5csRFRUFX1/fPP8jEBcXh6FDh2Lw4MHYtWsXxo4d+8zfC5kgQUTF9tVXXwkAYuXKlUIIIZKTk0XFihXFCy+8kKctADFz5kzddlhYmAAgwsLCnvgZI0aMEBYWFuLChQsFtomOjhYAxLp16wpsk5WVJTIyMoS3t7d45513dPsDAwNF06ZNn1hDxYoVRXBwcIHPP3z4UDg5OYmePXvq7c/OzhZNmjQRrVq1KvR7FWTo0KGiZs2auu1OnTqJN954Q1SqVEls2LBBCCHEb7/9JgCIvXv36tr5+fkJPz8/3fbt27fz/Ba5Zs6cKQCIBQsW6O0fO3assLKyEjk5OQXWl5iYKKytrUX37t319l+/fl2o1WoxePBg3b6goCABQHz33Xd6bbt37y7q1q2r2/7mm28EALFlyxa9didOnBAAxPLlywus59/Hc/v27QLb1KtXTzRr1kxkZmbq7Q8MDBRubm4iOzs739dlZWWJzMxMERAQIPr06aPbn/tvsVatWiIjIyPfep72/R49elQAEJ9++qleu5iYGGFtbS2mTJmi2+fn5ycAiAMHDjzhmyB6OvYQET2DNWvWwNraGgMHDgQAVKxYEa+88goOHz6cp0eguHbv3o0OHTqgfv36RXpdVlYWQkJC0KBBA1haWsLc3ByWlpa4dOmS3imcVq1a4cyZMxg7dix++eUXJCUl5XmvVq1aYf369ZgzZw6OHTuGzMxMveePHDmCe/fuISgoCFlZWbpbTk4OXnzxRZw4cQIPHz4s1HsVJCAgAFeuXEF0dDTS0tIQERGBF198ER06dMC+ffsAyF4jtVqNdu3aFem7elyvXr30ths3boy0tDQkJCQU+JqjR48iNTU1z+k3Dw8PdOzYMc+pQ5VKhZ49e+b5nGvXrum2f/75Zzg6OqJnz55632vTpk2h0Wie+QrFy5cv46+//sKQIUMAQO8zunfvjri4OL3exJUrV+K5556DlZUVzM3NYWFhgQMHDuQ5JQjI79DCwiLfz33a9/vzzz9DpVJh6NChejVpNBo0adIkz3FXqlQJHTt2fJavgoinzIiK6/Llyzh06BB69OgBIQTu37+P+/fvo1+/fgAeXXn2rG7fvo1q1aoV+XUTJ07EjBkz0Lt3b+zYsQPHjx/HiRMn0KRJE71TLdOmTcMnn3yCY8eOoVu3bnB2dkZAQABOnjypa/Ptt98iKCgIq1evRps2beDk5IRhw4YhPj4eAHSnMPr16wcLCwu92/z58yGE0J1yedp7FST3NNj+/fsRERGBzMxMdOzYEZ06ddKFjf3796Nt27awtrYu8vf1b87OznrbarUaAJ54iir3lJ+bm1ue59zd3fOcErSxsYGVlVWez0lLS9Nt37p1C/fv34elpWWe7zU+Pv6ZL6fP/d0mT56c5/1zTzvlfsbChQsxZswYtG7dGlu2bMGxY8dw4sQJvPjii/l+L/l9D7me9v3eunULQgi4urrmqevYsWN5jvtJn0VUWLzsnqiY1q5dCyEEfvjhB/zwww95nt+wYQPmzJkDMzOzZ/qcKlWq4MaNG0V+3aZNmzBs2DCEhITo7b9z5w4cHR112+bm5pg4cSImTpyI+/fvY//+/Zg+fTq6du2KmJgY2NjYoHLlyli8eDEWL16M69ev46effsLUqVORkJCAPXv2oHLlygCAJUuW4Pnnn8+3HldXVwB46nsVpFq1aqhTpw7279+PGjVqoEWLFnB0dERAQADGjh2L48eP49ixY7pL6sta7h/5uLi4PM/FxsbqvqOiqFy5MpydnQv8Xuzs7Ir8no+/PyBDcd++ffNtU7duXQDy35O/vz9WrFih93xycnK+r1OpVM9Ul0qlwuHDh3Vh6d8e3/csn0WUi4GIqBiys7OxYcMG1KpVC6tXr87z/M8//4xPP/0Uu3fvRmBg4DN9Vrdu3bBx40ZcvHhR98epMFQqVZ4/HDt37sTNmzdRu3btfF/j6OiIfv364ebNmwgODsbVq1fzXLJevXp1jB8/HgcOHMBvv/0GAGjbti0cHR1x4cIFjB8/vtA15vdeT9KpUyd899138PDw0E1zUKdOHVSvXh3vv/8+MjMzCxxQnaswvT3F0aZNG1hbW2PTpk145ZVXdPtv3LiBgwcP6noOiyIwMBChoaHIzs4ulbmt6tatC29vb5w5cyZPcH5cfv+ezp49i6NHj8LDw6NE6woMDMS8efNw8+ZN9O/fv0Tfm6ggDERExbB7927ExsZi/vz5+U7u5+Pjg6VLl2LNmjXPHIg++OAD7N69G+3bt8f06dPRqFEj3L9/H3v27MHEiRP1rt75t8DAQKxfvx716tVD48aNERkZiY8//jjP6beePXvCx8cHLVq0QJUqVXDt2jUsXrwYnp6e8Pb2hlarRYcOHTB48GDUq1cPdnZ2OHHiBPbs2aPrVahYsSKWLFmCoKAg3Lt3D/369YOLiwtu376NM2fO4Pbt21ixYkWh3utJAgICsHz5cty5cweLFy/W279u3TpUqlRJ75L7/NjZ2cHT0xPbt29HQEAAnJycULlyZdSoUeOpn/8kjo6OmDFjBqZPn45hw4Zh0KBBuHv3LmbPng0rKyvMnDmzyO85cOBAfP311+jevTvefvtttGrVChYWFrhx4wbCwsLw0ksvoU+fPk99nx07duTbm9SvXz988cUX6NatG7p27Yrhw4ejatWquHfvHv7880/88ccfuismAwMD8eGHH2LmzJnw8/PDxYsX8cEHH8DLywtZWVlFPrYnadu2Ld5880289tprOHnyJNq3bw9bW1vExcUhIiICjRo1wpgxY0r0M4l4lRlRMfTu3VtYWlqKhISEAtsMHDhQmJubi/j4eCFE8a8yE0JeXTNixAih0WiEhYWFcHd3F/379xe3bt0SQuR/lVliYqIYOXKkcHFxETY2NqJdu3bi8OHDea66+vTTT4Wvr6+oXLmysLS0FNWrVxcjR44UV69eFUIIkZaWJkaPHi0aN24s7O3thbW1tahbt66YOXOmePjwoV6d4eHhokePHsLJyUlYWFiIqlWrih49eojvv/++yO+Vn8TERFGhQgVha2urdwXT119/LQCIvn375nnN48crhBD79+8XzZo1E2q1WgAQQUFBQoiCr8pat26dACCio6OfWuPq1atF48aNhaWlpXBwcBAvvfSSiIqK0msTFBQkbG1t87w29/P/LTMzU3zyySeiSZMmwsrKSlSsWFHUq1dPjBo1Sly6dOmJteS+X0G3XGfOnBH9+/cXLi4uwsLCQmg0GtGxY0fd1ZNCCJGeni4mT54sqlatKqysrMRzzz0nfvzxRxEUFCQ8PT117XL/LX788ccF1lPY73ft2rWidevWwtbWVlhbW4tatWqJYcOGiZMnT+ra+Pn5iYYNGz7xeyAqDJUQQpR1CCMiIiIyJLzKjIiIiEweAxERERGZPAYiIiIiMnkMRERERGTyGIiIiIjI5DEQERERkcnjxIyFlJOTg9jYWNjZ2XGaeCIionJCCIHk5GS4u7ujQoWC+4EYiAopNja2xKenJyIiorIRExPzxIWyGYgKKXfa+5iYGNjb2ytcDRERERVGUlISPDw8nroYMgNRIeWeJrO3t2cgIiIiKmeeNtyFg6qJiIjI5DEQERERkcljICIiIiKTxzFEJSw7OxuZmZlKl6EICwsLmJmZKV0GERFRkTEQlRAhBOLj43H//n2lS1GUo6MjNBoN52oiIqJyhYGohOSGIRcXF9jY2JhcIBBCICUlBQkJCQAANzc3hSsiIiIqPAaiEpCdna0LQ87OzkqXoxhra2sAQEJCAlxcXHj6jIiIyg0Oqi4BuWOGbGxsFK5EebnfgamOoyIiovKJgagEmdppsvzwOyAiovKIgYiIiIhMHgMRERERmTwGIhOlUqmeeBs+fDgAICwsDB06dICTkxNsbGzg7e2NoKAgZGVlKXsAREREJYiByETFxcXpbosXL4a9vb3evs8++wxRUVHo1q0bWrZsiUOHDuHcuXNYsmQJLCwskJOTo/QhEBGRscjKAg4cULQEXnZvojQaje6xg4MDVCqV3j4AWLduHdzc3LBgwQLdvlq1auHFF18sszqJiMjILV8OzJ0L3LgBnDoFNG2qSBkMRKVBCCAlRZnPtrEBSuhKL41Gg7i4OBw6dAjt27cvkfckIiLSWbcOGDdOPq5SBbh6lYHIqKSkABUrKvPZDx4AtrYl8lavvPIKfvnlF/j5+UGj0eD5559HQEAAhg0bBnt7+xL5DCIiMlGRkcCYMfLxlCnABx8AarVi5XAMERXIzMwM69atw40bN7BgwQK4u7vjo48+QsOGDREXF6d0eUREVF7dvg307QukpwM9e8pTZgqGIYCBqHTY2MieGiVupTBbdtWqVfHqq69i2bJluHDhAtLS0rBy5coS/xwiIjJyDx8C27cDzz8PXL8O1K4NfPUVUEH5OMJTZqVBpSqx01aGplKlSnBzc8PDhw+VLoWIiMqT1FSgVSvgwgW57eUF7NgBODoqWlYuBiIq0BdffIHTp0+jT58+qFWrFtLS0vDVV18hKioKS5YsUbo8IiIqT1atkmGoUiVg6FBg1izAyUnpqnQYiKhArVq1QkREBEaPHo3Y2FhUrFgRDRs2xI8//gg/Pz+lyyMiIkOWkgIcPw7Ex8tL6j/5RO6fNw94801la8uHSgghlC6iPEhKSoKDgwO0Wm2eK6zS0tIQHR0NLy8vWFlZKVShYeB3QUREuHdPnh773//099esCfz5J2BpWWalPOnv97+xh4iIiIhKjhDycvr//U+eEmvSBNBogIYNgREjyjQMFQUDEREREZWMP/8E3n8f+OEHwMwM2LMHaNlS6aoKhYGIiIiIik+rBVaulJfP515BZmEh95WTMAQwEBEREdHjMjOBL74AzpwBrKyAunUBDw9g717gyhXAxUXOLq3VAr17y4kWARmEOnWSvUTPP6/oIRQVA1EJ4vh0fgdEROVedDQwYABw4sST23311aPHderIgPTyywYzr1BRMRCVAAsLCwBASkoKrK2tFa5GWSn/LGqb+50QEZGBysmR64kJAdSrJy+L//57IDYWSEqSwWbcOCArS44N+t//gGbNgI4dgZ9+An78Ub5Pnz7Axo3lfkJiBqISYGZmBkdHRyQkJAAAbGxsoCqhFefLCyEEUlJSkJCQAEdHR5iZmSldEhERFUSrBQYNAnbvltsVKsiAlKtlS2DLFnmaLD+vvSaX4UhLA5ydS7/eMsBAVEI0Gg0A6EKRqXJ0dNR9F0REZICysuSCqocPywVVnZyAuDg5/mfRIqB+faBt26cvtmprW+57hf6NgaiEqFQquLm5wcXFBZmZmUqXowgLCwv2DBERGbL//Q+YP1+GITs7ICwMeO45ICpKBiBvb6UrVAwDUQkzMzNjKCAiIsPz9ddAUBCQnS23V60CmjeXj318lKvLQFRQugAiIiIqZatXA6++KsPQCy8AmzcDAwcqXZVBYQ8RERFRaUhLkyHE1RVYvBgwV+hP7rp1wBtvyMdjxgBLl8pB1KSH3wgREVFp+OwzuYTFsmVyDa/k5NL7LCGAfv0Ae3ugVy/g9Gm5/8YN4K235ON33pG1MAzli6vdF1JhV8slIiIjk50NTJgAbN8ODB8uFyu1swNq1ZITEgoB/P034OYGxMQAs2fLthkZ+u9TpYoct/PWW3KB07FjgddfB27eBGrXlld5Fde2bUDfvo+2zc0BPz8ZiC5eBNq0ASIiTDIMFfrvt1BQeHi4CAwMFG5ubgKA2LZtm97zOTk5YubMmcLNzU1YWVkJPz8/cf78eb02aWlpYvz48cLZ2VnY2NiInj17ipiYGL029+7dE0OHDhX29vbC3t5eDB06VCQmJhapVq1WKwAIrVZbnEMlIqLyKihICBl78t6qVRPCw0M+trYWQqXSf75lSyF27BDC2zv/1+e2b9RIiLNnH31mYqIQb78txNq1QiQkCHHqlBCZmXJ/dvajdt99J0TVqo/er2dPIfr00f8Ma2shTp8u06/MkBT277eiY4gePnyIJk2a4LXXXsPLL7+c5/kFCxZg4cKFWL9+PerUqYM5c+agc+fOuHjxIuzs7AAAwcHB2LFjB0JDQ+Hs7IxJkyYhMDAQkZGRuqu9Bg8ejBs3bmDPnj0AgDfffBOvvvoqduzYUXYHS0REhi87G7h0Sfb0JCXJfRs2yB6X996Ta3slJsrnzp+XPTCAXNk9NVU+fvllOcNzbCwQEABoNEC7dvJKrps3ZZt33pGn02JiAJUKOHdOrv31n/8A6enAL78Ap07p12ZpKXudqlcHunaVn507sSIgJ1HcvFnODfTHH7I+Cws5p5CnZ+l+b8agjALaU+GxHqKcnByh0WjEvHnzdPvS0tKEg4ODWLlypRBCiPv37wsLCwsRGhqqa3Pz5k1RoUIFsWfPHiGEEBcuXBAAxLFjx3Rtjh49KgCIv/76q9D1sYeIiMjIpacL0a5d/j05b7+dt71WK8Rvvwmxf78QDx4IERkpxMWLBb9/WJgQGo0QCxfK7dRUIf7+W4i4OCE6d877mU5OQqjV8rGtbcG9VBMnyh6ku3dL4Usp/8pFD9GTREdHIz4+Hl26dNHtU6vV8PPzw5EjRzBq1ChERkYiMzNTr427uzt8fHxw5MgRdO3aFUePHoWDgwNat26ta/P888/DwcEBR44cQd26dcv0uIiIyEAtXSrH2VhayjE9iYlyBmc7O9k79Dh7e8DX99H2c889+f39/eX75bKyejQR4q5dwIIFslfIzU327IwaBdjYPOoVunIFqFxZrjh/8aJ8rmdPuQ4ZPTODDUTx8fEAAFdXV739rq6uuHbtmq6NpaUlKlWqlKdN7uvj4+Ph4uKS5/1dXFx0bfKTnp6O9PR03XZSbtcpEREZn2PHgFmz5OMVK+RVYUlJ8qqsNm3kgOjSZG4OTJ/+5DZ16sh7zh9UKgw2EOV6fJFUIcRTF059vE1+7Z/2PnPnzsXs2bOLWC0REZUbf/0le1pu3ZJXkWVkyPAzfLh83t4emDZN0RKp7Bjs9Xe5C4Q+3ouTkJCg6zXSaDTIyMhAYmLiE9vcunUrz/vfvn07T+/Tv02bNg1arVZ3i4mJeabjISIihSQkAPfv6++7e1cONu7dW56aysgA+vSRg5lN8NJ0MuBA5OXlBY1Gg3379un2ZWRkIDw8HL7/nLNt3rw5LCws9NrExcXh/PnzujZt2rSBVqvF77//rmtz/PhxaLVaXZv8qNVq2Nvb692IiKgcuX4dWL5cjr+pUQP47jt5BdcXXwAtWgD37sl2FSoAH3wgr/r65wpmMj2KnjJ78OABLl++rNuOjo7G6dOn4eTkhOrVqyM4OBghISHw9vaGt7c3QkJCYGNjg8GDBwMAHBwcMHLkSEyaNAnOzs5wcnLC5MmT0ahRI3Tq1AkAUL9+fbz44ot444038MUXXwCQl90HBgZyQDURkTF6+FD2+nz99aN96enAgAEyCJ08+Wj/wYNAq1byUnUyaYoGopMnT6JDhw667YkTJwIAgoKCsH79ekyZMgWpqakYO3YsEhMT0bp1a+zdu1c3BxEALFq0CObm5ujfvz9SU1MREBCA9evX6604//XXX+Ott97SXY3Wq1cvLF26tIyOkoiISs3163KZir/+AiIj5WmwzZuB48dlz0/t2nI26Bs3gM8/l2HIzExeHdahg7wRgUt3FBqX7iAiMiB//QUMGwacOJH/805OcjmL9u3lthDAm2/KhU4XLZKDqMkkFPbvNwNRITEQEREZgIcP5ZxAq1bJmaHNzOQM0LVry9mYly2T8/scOAA0b5739UlJ8uoxMhmF/ftt8JfdExERISNDztXz1lvA2rVyX8eOwMaNgLv7o3bvvitPlVWunP/7MAxRARiIiIjIcD14IMcAbd8OpKU92r9li7xM/vH55PKZiJeoMBiIiIjIcE2eDHz7rf6+IUOAvn2VqYeMFgMREREZlsREYOtWYMcO2TMEyPurV4HffwcWLlS0PDJODERERFQ2rl0DwsIAtRp4/nk5WeKNG3Lcz9698pL5/v3l9s2bj143dSrQq5diZZNpYCAiIqLSER8P7N4N/PGHHBD95ZfyKrFcjo6AVisviQeAn36SNwDw8pJrivXsCTRrVtaVkwliICIiopL34Ydy9ficHP39jRsDNjZygsTc9cX8/IBJk4D9+2Vo8vaWPUZPWG+SqKRxHqJC4jxERESFdO4c0LSpDEMtWwIvvCB7hurXB8aNk71FqanApUuAszNQteqj16any7mFzPn/61QyOA8RERGVnjt3gA0bgLp1gR49Hl3+LoS8MiwnB3jlFbmgan6srWVv0ePU6tKrmegJGIiIiKhwhAA++kiuCXbvHpCdLfd36SIvjXd0lEtj7N0LWFgAc+cqWi5RUTAQERFRwTIz5emrCxeA+fPlzNC5GjWSp7327pVXjdWrB/zyi3zuww+BWrWUqZmoGCooXQARERmo998HLC0BOzu5XlhuGFq8WF4uf+YMcOyYnB364sVHs0l37SpPmxGVI+whIiKivA4fBubMkY8fPpTBKDfo5K4gDwBNmsjL6n/6CcjKkpfIt2kjB0YTlSMMRERE9IgQwJQpcpyQEEBQEPCf/8h5gWxs8n9N1arAmDFlWydRCWMgIiKiRw4dAj75RD7295fBiFONkAngGCIiInokNFTeDxsml9lgGCITwUBERERSVhbwww/y8dChytZCVMYYiIiICIiJkeOF7twBqlQBOnRQuiKiMsUxREREpujSJWDGDDnPkI0N8OOPwIMH8rkpU7h0Bpkc/osnIjI1p04BAQFAYqL+/jZtgEWLgNatlamLSEEMREREpmbGDBmGWrUChgyRC6rWrAn06QNU4EgKMk0MREREpuTWLWDPHvn4q6/k4qxExEHVREQmZf16uShr69YMQ0T/wh4iIiJTkJMDDBwIfP+93A4KUrYeIgPDQEREZAp++02GITMzYNAg4LXXlK6IyKAwEBERmYJNm+T9sGHA2rXK1kJkgDiGiIjIWN2+DezfD6SlPTpVNmSIsjURGSj2EBERGZvUVODyZaBHDzkDdfXq8jJ7Nze5YCsR5cEeIiIiY5GZCXz0EeDsDDRuLMMQAFy/DlhaAsuWyTFERJQHAxERkTEQAhg1Cvjvf2UPEQC0aAFs3Qq8+ipw8KCceJGI8sVTZkRE5VVODnDsGPDxx8Dp08DVq3Km6TVr5GX1KpVsxyBE9FQMRERE5UVOjrx8/rPPgF9+kRMs5vYG5Vq4EBg+XJHyiMozBiIiovLg2jWga1fg4kX9/XZ2sgcoKAioVg2oU0eZ+ojKOQYiIiJDd/cu0LkzcOkSYG8P9O0LjB0LODoCNWoAFhZKV0hU7jEQEREZuhkzZBjy9AQiImRPEBGVKF5lRkRkyC5dAlatko83bGAYIiolDERERIYoPR349lugY0c5eLp7d8DPT+mqiIwWT5kRERmamBg5o/SVK3K7Zk15ZRkRlRr2EBERGQohgMhI2Rt05Qrg6irHD509C9SurXR1REaNPURERIZg9Wrg/feBuDi5rdEAx4/LdciIqNSxh4iISGkXLwJjxsgwZGkJDB4MHDrEMERUhthDRESkpIwMIDgYyMoCunUDtmwBrK2VrorI5LCHiIhICUIAu3YBLVsCe/YA5uZy2Q2GISJFMBAREZWV9HS5Hhkgxwv16CEHTFeuDPzwA1CvnrL1EZkwnjIjIipNf/0lQ0/16nItsooV5SX1mzfL54ODgenTgSpVlKySyOQxEBERlaTjx4F33gEePpQTKkZFyf1mZnI7KelRGBo3Dli0SLlaiUiHgYiIqKRcuQIEBgJ37ujvt7IC0tLkWmQLFsiryqysgLfeUqZOIsqDgYiIqKRMmCDD0HPPASEhQIUKckLFlBS5HtmoUUCDBkpXSUT5UAkhhNJFlAdJSUlwcHCAVquFvb290uUQkaGJjgZq1ZJXj128CNSpo3RFRITC//3mVWZERCVh1SoZhjp3ZhgiKod4yoyIqDi0Wjlv0O7dgEoFnDgh948erWxdRFQsBt1DlJWVhf/+97/w8vKCtbU1atasiQ8++AA5ufN4ABBCYNasWXB3d4e1tTX8/f0RlXtVxz/S09MxYcIEVK5cGba2tujVqxdu3LhR1odDROVd7giDnBygd2/ggw9kEPr9d/lcUBDQq5eiJRJR8Rh0IJo/fz5WrlyJpUuX4s8//8SCBQvw8ccfY8mSJbo2CxYswMKFC7F06VKcOHECGo0GnTt3RnJysq5NcHAwtm3bhtDQUERERODBgwcIDAxEdna2EodFROVNZibwySdyAsXu3eXcQb/+CtjYAOvWAaGhwJkzwPr1csZpIip3DHpQdWBgIFxdXbFmzRrdvpdffhk2NjbYuHEjhBBwd3dHcHAw3n33XQCyN8jV1RXz58/HqFGjoNVqUaVKFWzcuBEDBgwAAMTGxsLDwwO7du1C165dC1ULB1UTmaijR+XVYefO5X1u0SIZjojIYBnFoOp27drhwIED+PvvvwEAZ86cQUREBLp37w4AiI6ORnx8PLp06aJ7jVqthp+fH44cOQIAiIyMRGZmpl4bd3d3+Pj46NrkJz09HUlJSXo3IjIRaWnA/Pny9JevrwxDzs7AkiUyHL38MvD55/IyeyIyCgbdt/vuu+9Cq9WiXr16MDMzQ3Z2Nj766CMMGjQIABAfHw8AcHV11Xudq6srrl27pmtjaWmJSpUq5WmT+/r8zJ07F7Nnzy7JwyGi8kAI4I03gE2bHu177TU5oWLlysrVRUSlyqB7iL799lts2rQJmzdvxh9//IENGzbgk08+wYYNG/TaqVQqvW0hRJ59j3tam2nTpkGr1epuMTExxT8QIio/FiyQYcjMDJgzRw6YXruWYYjIyBl0D9F//vMfTJ06FQMHDgQANGrUCNeuXcPcuXMRFBQEjUYDQPYCubm56V6XkJCg6zXSaDTIyMhAYmKiXi9RQkICfH19C/xstVoNtVpdGodFRIbqp5+AadPk488/B8aOVbYeIiozBt1DlJKSggoV9Es0MzPTXXbv5eUFjUaDffv26Z7PyMhAeHi4Luw0b94cFhYWem3i4uJw/vz5JwYiIjJid+8C58/LVehXrgS2bgWWLgVeeUWeMhs9mmGIyMQYdA9Rz5498dFHH6F69epo2LAhTp06hYULF2LEiBEA5Kmy4OBghISEwNvbG97e3ggJCYGNjQ0GDx4MAHBwcMDIkSMxadIkODs7w8nJCZMnT0ajRo3QqVMnJQ+PiMrarVtAcjLQrp18nJ/cAdNEZFIMOhAtWbIEM2bMwNixY5GQkAB3d3eMGjUK77//vq7NlClTkJqairFjxyIxMRGtW7fG3r17YWdnp2uzaNEimJubo3///khNTUVAQADWr18PMzMzJQ6LiJSwaBEwcaKcVVoIOUbIzAx44QUZkqysgA4dgP/+l3MJEZkgg56HyJBwHiKicmrPHuCHH+TA6Nz/3FWtChw/Dri5yRXpichoFfbvN/83iIiMV2ysXGIjPV1uv/GGvIS+Vi3AxUXR0ojIsDAQEZHxWrBAhqHGjeXVY/368XQYEeWL/2UgIuNw/Trw5pvApUtyrNCDB/IKMkCuQ9a5s7L1EZFBYyAiovLv77+BNm2Ae/fk9vjxj57r0QPgFaVE9BQMRERUPj18CMTFAVWqyDXH7t0DmjaVp8V27wasrYEBA+SYoafMXE9ExEBEROVHUpJcULVFC3kZ/dWr8vHFi0C1ajIIaTTAe+8pXSkRlTMMRERUfqxeDXz1lbzlOnFC3m/eLMMQEVExcAIOIio/vv/+0WMHB6BlS/l43Dg5wSIRUTGxh4iIyoeYGODYMTke6OhRwNMTqFgR2L9fDpwmInoGDEREVD6sXy/v27UDWrd+tL93byWqISIjw1NmRGT4fv0V+OAD+XjkSEVLISLjxEBERIbr9m1gyhQ5qWJWFjBwIDBsmNJVEZER4ikzIjI8ycnA++8DX3wBpKbKfS+/LK8y45xCRFQKGIiIyLAIAQwZAuzYIbdbtABmzgQCA5Wti4iMGgMRERmWlStlGLK0BLZskVeQsVeIiEoZxxARkbKEAKKj5RihBw+AGTPk/gULZK8QwxARlQEGIiJSzv37QP/+QM2a8jZgAHD3LuDtrb9AKxFRKeMpMyJSRmamPB125IjcjomRNwCYOhUwM1OuNiIyOQxERKSM6dNlGHJwAH7+Gbh2TV5F5uAADB2qdHVEZGJUQgihdBHlQVJSEhwcHKDVamFvb690OUTlm1YLVKkie4m2bgX69FG6IiIyUoX9+80xRERU9n75RYahunUZhojIIDAQEVHZy51jqGdPZesgIvoHAxERla2sLGDnTvmYgYiIDAQDERGVrYMHgcREwMkJ8PVVuhoiIgAMRERU1taskfeDBgHmvNCViAwDAxERlZ07d4Bt2+Tj119XthYion/h/54RUekTAsjJkctyZGYCzz0HNG2qdFVERDoMRERUujIzgRdfBMLDgexsuTbZnDlKV0VEpIeBiIhK1/vvy4HUuebMAbp1U64eIqJ8MBARUek5dQqYP18+Xr0aaNECaNJE2ZqIiPLBQEREpee99+T4oQEDgJEjla6GiKhADEREVHJSU4Hdu4HvvwfCwoBbt+Sl9RwzREQGjoGIiErGrVvA888DV6/q7w8OBmrXVqIiIqJCYyAiomcnhJxX6OpVwNUVGDoU6N0b8PaW20REBo6BiIieTVycHB+0ezdgaQns3w/4+ChdFRFRkTAQEVHxXLkCHD4sL6u/fh1Qq4FVqxiGiKhcYiAioqILC5OTLWZkyO06dYAffwTq11e0LCKi4mIgIqKnS0oCtmyRp8dq1QLefFOGoUaNgHbtgA8/BJydla6SiKjYGIiI6MkuXAA6dwZiY/X3+/vLcUNWVoqURURUkrjaPRHllZAAjBkje30aN5ZhqEYN2RsEAEFBDENEZFTYQ0REeQ0apL/+WJs2wM8/A05OQGIiUKmScrUREZUCBiIi0nf+vAxDZmbAzp1yoLSHh1ylHmAYIiKjxEBERI8IAXz+uXz80ktA167K1kNEVEYYiIhIunYNePllIDJSbo8Zo2w9RERliIOqiUxdaiqwbBnwwgsyDNnYANOnAwEBSldGRFRm2ENEZOoGDQK2b5eP69YF9u2TY4aIiEwIe4iITFlEhAxDZmbA3LnAkSMMQ0RkkthDRGSKHjwAunUDTpyQ26+/DkydqmxNREQKYiAiMkWrVsneIUBOvjhjhrL1EBEpjKfMiExJfDywfDkwaZLcXrgQiIkBqlZVti4iIoWxh4jIVJw+DfTo8WhNMnd3YOxYQK1WtCwiIkNg8D1EN2/exNChQ+Hs7AwbGxs0bdoUkbnzpAAQQmDWrFlwd3eHtbU1/P39ERUVpfce6enpmDBhAipXrgxbW1v06tULN27cKOtDIVLOH38AHTrIMFSzJuDnB6xYwTBERPSPYgWiESNGIDk5Oc/+hw8fYsSIEc9cVK7ExES0bdsWFhYW2L17Ny5cuIBPP/0Ujo6OujYLFizAwoULsXTpUpw4cQIajQadO3fWqy84OBjbtm1DaGgoIiIi8ODBAwQGBiI7O7vEaiUyWEeOyNXq798HfH3lXEO//gr06qV0ZUREBkMlhBBFfZGZmRni4uLg4uKit//OnTvQaDTIysoqkeKmTp2K3377DYcPH873eSEE3N3dERwcjHfffReA7A1ydXXF/PnzMWrUKGi1WlSpUgUbN27EgAEDAACxsbHw8PDArl270LWQSxMkJSXBwcEBWq0W9vb2JXJ8RKXu5Ek54WJaGtC6NfDLL4CDg9JVERGVmcL+/S5SD1FSUhK0Wi2EEEhOTkZSUpLulpiYiF27duUJSc/ip59+QosWLfDKK6/AxcUFzZo1w5dffql7Pjo6GvHx8ejSpYtun1qthp+fH44cOQIAiIyMRGZmpl4bd3d3+Pj46NrkJz09Xe/4kpKSSuy4iMpEdrZcfiMtDejSBThwgGGIiKgARRpU7ejoCJVKBZVKhTp16uR5XqVSYfbs2SVW3JUrV7BixQpMnDgR06dPx++//4633noLarUaw4YNQ3x8PADA1dVV73Wurq64du0aACA+Ph6Wlpao9NgK3a6urrrX52fu3LkleixEZSolBZg2TfYQ2dsDGzYAtrZKV0VEZLCKFIjCwsIghEDHjh2xZcsWODk56Z6ztLSEp6cn3N3dS6y4nJwctGjRAiEhIQCAZs2aISoqCitWrMCwYcN07VQqld7rhBB59j3uaW2mTZuGiRMn6raTkpLgwRl8yRDdvQtcvQo89xygUsmB0926AWfPyufnzgU0GkVLJCIydEUKRH5+fgDkqarq1as/NXQ8Kzc3NzRo0EBvX/369bFlyxYAgOaf/8jHx8fDzc1N1yYhIUHXa6TRaJCRkYHExES9XqKEhAT4+voW+NlqtRpqXoFDhiQ+Xi7E6uUlt1NSgMWLgXnzgORkIDBQ3ubMAW7cAFxdgZUrgZdeUrRsIqLyoFhXmXl6eiIiIgJDhw6Fr68vbt68CQDYuHEjInJnvy0Bbdu2xcWLF/X2/f333/D09AQAeHl5QaPRYN++fbrnMzIyEB4ergs7zZs3h4WFhV6buLg4nD9//omBiMigHDwI1KkDeHvL0OPgADg6Au+9J8MQAPz8MzB6tAxDdesCR48CvXvLXiMiInqiYgWiLVu2oGvXrrC2tsYff/yB9PR0AEBycrLu9FZJeOedd3Ds2DGEhITg8uXL2Lx5M1atWoVx48YBkKfKgoODERISgm3btuH8+fMYPnw4bGxsMHjwYACAg4MDRo4ciUmTJuHAgQM4deoUhg4dikaNGqFTp04lVitRqfnhB3kKLDlZDpTeuRNISgIyM4EaNYDNm4Hz54E33gBatQImT5aX1uf2JBER0dOJYmjatKnYsGGDEEKIihUriv/9739CCCFOnTolXF1di/OWBdqxY4fw8fERarVa1KtXT6xatUrv+ZycHDFz5kyh0WiEWq0W7du3F+fOndNrk5qaKsaPHy+cnJyEtbW1CAwMFNevXy9SHVqtVgAQWq32mY+J6Kk2bhRi7FghhgwRQqUSAhCib18hvv5aiJEjhQgLE+LKFSEyM5WulIjIoBX273ex5iGysbHBhQsXUKNGDdjZ2eHMmTOoWbMmrly5ggYNGiAtLa3kk5vCOA8RlZmvvgKCgvT3jRkDLFkCmJkpUxMRUTlV2L/fxVrLzM3NDZcvX0aNGjX09kdERKBmzZrFeUsiAoBLl4DXX5ePX3oJ8PQEhg4FWrZUti4iIiNXrEA0atQovP3221i7di1UKhViY2Nx9OhRTJ48Ge+//35J10hkOhYskGODOncGtm4FKhj8coNEREahWIFoypQp0Gq16NChA9LS0tC+fXuo1WpMnjwZ48ePL+kaiYxbcrK8MiwlRZ4uA4CZMxmGiIjKUJHHEGVnZyMiIgKNGjWClZUVLly4gJycHDRo0AAVK1YsrToVxzFEVCoyMoDmzeVVYrnatQMKWL+PiIiKptTGEJmZmaFr1674888/4eTkhBYtWjxToUQmSwg5seK/w1DfvsCiRYqVRERkqop1yqxRo0a4cuUKvDjPCVHxnD0LdOgA3Lsnt9esAQYPBqyslK2LiMhEFWuQwkcffYTJkyfj559/RlxcHFeFJyqqjz56FIa6dweGD2cYIiJSULHmIarwr8Ge/17PTPyzYGp2dnbJVGdAOIaISsz160DNmnLW6VOngKZNla6IiMholeo8RGFhYcUujMikZWYCb78tw1DHjgxDREQGoliBKHfVeyIqgpwcOU7oxx8BCwtg1iylKyIion8UKxCdPXs23/0qlQpWVlaoXr061Gr1MxVGZFSEkHML/fADYGkpQ9ELLyhdFRER/aNYgahp06Z6Y4ceZ2FhgQEDBuCLL76AFQeKkqnKygKiooAHD4Bly4BvvpH7V62Sq9cTEZHBKNZVZtu2bYO3tzdWrVqF06dP49SpU1i1ahXq1q2LzZs3Y82aNTh48CD++9//lnS9ROVDXBzQrJkcI9SunQxDZmbAxx/nXbiViIgUV6weoo8++gifffYZunbtqtvXuHFjVKtWDTNmzMDvv/8OW1tbTJo0CZ988kmJFUtULty5A/j5yYVara0Be3ugfXvgnXeANm2Uro6IiPJRrEB07tw5eHp65tnv6emJc+fOAZCn1eLi4p6tOqLyJiMD6NNHhiFPT+DgQXmJPRERGbRinTKrV68e5s2bh4yMDN2+zMxMzJs3D/Xq1QMA3Lx5E66uriVTJVF5MWsWEBEhe4V27WIYIiIqJ4rVQ7Rs2TL06tUL1apVQ+PGjaFSqXD27FlkZ2fj559/BgBcuXIFY8eOLdFiiQzagQPA/Pny8dq1QIMGytZDRESFVqyZqgHgwYMH2LRpE/7++28IIVCvXj0MHjwYdnZ2JV2jQeBM1VSg+Hjgiy/kchyZmcCQIcCmTUpXRUREKOWZqgGgYsWKGD16dHFfTmQcLl+WA6Xv3JHb/foBX36pbE1ERFRkxRpDBAAbN25Eu3bt4O7ujmvXrgEAFi1ahO3bt5dYcUQG6/59YNIkuWL9nTtA/fqyV+jbb+WVZUREVK4UKxCtWLECEydORLdu3ZCYmKhbzLVSpUpYvHhxSdZHZJjGjAEWLgRu3ABq1JBXkw0ZAlQo9v9jEBGRgor1X+8lS5bgyy+/xHvvvQdz80dn3Vq0aKG77J7IaB0/DoSGAiqVHDt06hSg0ShdFRERPYNijSGKjo5Gs2bN8uxXq9V4+PDhMxdFZLB++w3o21c+fu014M03la2HiIhKRLF6iLy8vHD69Ok8+3fv3o369es/a01Ehik2FujeHUhIABo3BubNU7oiIiIqIcXqIfrPf/6DcePGIS0tDUII/P777/jmm28QEhKCNWvWlHSNRMrLzAQmTACSkoCWLYGwMMDWVumqiIiohBQrEL322mvIysrClClTkJKSgsGDB6Nq1apYsmQJXnjhhZKukahsCSEHTN++DTz3HKDVyhmoY2PlAq2rVjEMEREZmWJPzJjrzp07yMnJQXZ2NkJCQrB69WqkpqaWVH0GgxMzmpCffwZ69sy7X6MB5s4Fhg8v85KIiKh4Cvv3u0hjiO7fv48hQ4agSpUqcHd3x+effw4nJycsW7YMtWvXxrFjx7B27dpnLp5IMULIGacB4Pnn5emxmjWBTz4Brl1jGCIiMlJFOmU2ffp0HDp0CEFBQdizZw/eeecd7NmzB2lpadi1axf8/PxKq06ishEeDhw7BqjVwLZtvJyeiMhEFCkQ7dy5E+vWrUOnTp0wduxY1K5dG3Xq1OFkjGQ8QkLk/ciRDENERCakSKfMYmNj0eCfFbxr1qwJKysrvP7666VSGFGZO3kS2LdPDpz+z3+UroaIiMpQkQJRTk4OLCwsdNtmZmaw5dU2ZCxmzJD3Q4bI5TiIiMhkFOmUmRACw4cPh1qtBgCkpaVh9OjReULR1q1bS65CorKwaxewZw9gYfEoGBERkckoUiAKCgrS2x46dGiJFkOkmOnT5f3bbwO1aytbCxERlblnnofIVHAeIiN29izQpAlgaQnExQFOTkpXREREJaRU5iEiMkqbN8v7Hj0YhoiITBQDEZme778H2rSRty1bHgWiwYOVrYuIiBRTrLXMiMqtc+dk8MnKktv9+sl7Z2fZQ0RERCaJPURkOjIy5ISLWVlAt27AoEFyf5Uqcv0ya2tl6yMiIsWwh4hMgxDA2LHAiROAgwPw5ZeAmxvw6qtA06byMRERmSwGIjIN338PrFkDVKgAfPMNULWq3N+tm7J1ERGRQWAgIuN08SKQnQ3UqQOkpwOTJsn9//0vQxAREeXBQETGIz0d2L5d9gb98IPc5+YGeHoCN27I+6lTla2RiIgMEgMRGYecHNnzExYmtytUkIOk4+LkzcZGjhviwGkiIsoHrzKj8u/WLbn+WFgYYGsLjB4NREYCd+8Cn30GDB8OnDkDdO6sdKVERGSg2ENE5dtff8kJFu/fl9sffwyMGfPo+bfeUqQsIiIqX9hDRIZNCHl73OLFgLc3UL++DEM1agAffQSMGlXGBRIRkTFgDxEZtpdfBvbsAd58E5g8WY4B2rEDeOcd/Xa7dslwREREVAxc7b6QuNq9Ak6eBFq2fLRdoYIcPJ2rbl15ef1bb8mxQkRERI8p7N9v9hCR4fr8c3nfoYMMQuHhcrtmTeDFF4GFC4F79wCNRrkaiYjIKDAQkeERApg379Eq9PPny56i69cBR0fg3wmfS24QEVEJKFeDqufOnQuVSoXg4GDdPiEEZs2aBXd3d1hbW8Pf3x9RUVF6r0tPT8eECRNQuXJl2NraolevXrhx40YZV0+F9vXXwPTpcqbpYcMenTarXl0/DBEREZWQchOITpw4gVWrVqFx48Z6+xcsWICFCxdi6dKlOHHiBDQaDTp37ozk5GRdm+DgYGzbtg2hoaGIiIjAgwcPEBgYiOzs7LI+DCqM5cvl/ZQpwPr1ipZCRESmoVwEogcPHmDIkCH48ssvUalSJd1+IQQWL16M9957D3379oWPjw82bNiAlJQUbP7ndItWq8WaNWvw6aefolOnTmjWrBk2bdqEc+fOYf/+/UodEhXk3Dng6FHA3FxeSaZSKV0RERGZgHIRiMaNG4cePXqgU6dOevujo6MRHx+PLl266Pap1Wr4+fnhyJEjAIDIyEhkZmbqtXF3d4ePj4+uTX7S09ORlJSkd6MysG6dvO/dm4OliYiozBj8oOrQ0FBERkbi5MmTeZ6Lj48HALi6uurtd3V1xbVr13RtLC0t9XqWctvkvj4/c+fOxezZs5+1fCqqnTvl/aBBytZBREQmxaB7iGJiYvD222/j66+/hpWVVYHtVI+dVhFC5Nn3uKe1mTZtGrRare4WExNTtOKp6K5cAf7+W54uCwhQuhoiIjIhBh2IIiMjkZCQgObNm8Pc3Bzm5uYIDw/H559/DnNzc13P0OM9PQkJCbrnNBoNMjIykJiYWGCb/KjVatjb2+vdqJTt2SPvfX0BBwdlayEiIpNi0IEoICAA586dw+nTp3W3Fi1aYMiQITh9+jRq1qwJjUaDffv26V6TkZGB8PBw+Pr6AgCaN28OCwsLvTZxcXE4f/68rg0ZgKwsIDRUPu7WTdlaiIjI5Bj0GCI7Ozv4+Pjo7bO1tYWzs7Nuf3BwMEJCQuDt7Q1vb2+EhITAxsYGgwcPBgA4ODhg5MiRmDRpEpydneHk5ITJkyejUaNGeQZpUxlJTZWTL1pbA5GRwPffA7/9Jm/m5kDfvkpXSEREJsagA1FhTJkyBampqRg7diwSExPRunVr7N27F3Z2dro2ixYtgrm5Ofr374/U1FQEBARg/fr1MDMzU7ByE3XxItCpE3D7NuDhAVy+/Og5CwsZjurUUa4+IiIySVzctZC4uOsz2rkTGD8eiI0FMjIe7beykpfY+/rKNcse6xEkIiJ6FlzclQzHhQvAwIHAgwdy+7nngE8+Ae7elb1Fjo6KlkdERMRARKUrOxsYMkSGoQ4dgBUrAG9voIJBj+cnIiITw0BEpeuLL4DTp2UvUGgo4OKidEVERER58H/TqfTcvg289558/NFHDENERGSwGIio9EyfDty/DzRtCowapXQ1REREBWIgotLx66/AmjXy8dKlAKc4ICIiA8YxRFSy/v4b2LABWL9eTr74+utA27ZKV0VERPREDERUcnJygP79gTNn5HbdusDixYqWREREVBg8ZUbF98cfwFtvAX/9Jbe3bZNhyM4OePddORmjra2yNRIRERUCe4ioeKKjgS5d5OSKq1cD1arJfQDw9tvAhx8qWx8REVERMBBR0eXkyJmn794FbGyAlBTg0iX53HPPARMnKlsfERFRETEQUdHcuQPs3w/8/jtQsSIQFQVcvy5npPbykgu2qlRKV0lERFQkDERUeNu2Aa+8IsMPIHuCqleXNyIionKMg6qpcGJj5SX0uWHIxYWnxoiIyGgwENHTZWYCgwcD9+4BzZoBv/0GnDgBODgoXRkREVGJYCAifcuWyYHRQ4YA4eFARoZcdiM8XI4Z+vprwNeXp8mIiMioqIQQQukiyoOkpCQ4ODhAq9XC3t5e6XJKR0oKoNEAycmP9llbA6mp8vG2bUDv3oqURkREVByF/fvNHiJ6ZPv2R2Fo1CjAwkKGIUdH+RzDEBERGSkGInrkq6/k/YwZwMqVcp6hCxeAmBigVy9layMiIipFvOyepDNngL175eNXX5X3dnZA/frK1URERFRG2ENEclX68ePlDNSvvAJ4eytdERERUZliICJg1SogIkIuw/Hpp0pXQ0REVOZ4yswUCQGcOgUsXCiX4IiJkfvnzJFLbxAREZkYBiJjJwTQp4+cSLF3b8DZGVi//lEIyhUQIFepJyIiMkEMRMbul1/kJfMAsHz5o/02NkCPHnICxlu3gP79gQo8g0pERKaJgcjYzZ0r7+vUAV58UYafLl3kUhxWVsrWRkREZCAYiIzZ558Dhw7JCRYPHACqVVO6IiIiIoPEQGRMEhOBv/6Sp8i+/x64ckXunz2bYYiIiOgJGIiMRUwM0KgRoNU+2qdSAVOnyhsREREViIHIWEyfLsNQpUpA69bAG28A/v6Ak5PSlRERERk8BqLyTAh5f+QIsGmTfLxvH9C8uXI1ERERlUMMROVRRobsAdq+XQ6Yzr1c/rXXGIaIiIiKgYGoPPr440cr0+eqVw/47DNl6iEiIirnGIjKm4sXgQ8/lI9XrpRzCe3fD7z/vlydnoiIiIqMgag8yckB3nwTSE+Xkyy++aa8kiwoSOnKiIiIyjWu1VBefPyxXHj10CG57MaKFTIMERER0TNjD1F5EB0tL6vPygLMzIDFi4EaNZSuioiIyGgwEJUHH34ow1DHjsDmzYCrq9IVERERGRWeMjN0v/0GbNggH4eEMAwRERGVAgYiQ5acDLz6qhxM/eqrcgZqIiIiKnEMRIbsyy/l+CFPT2DJEqWrISIiMloMRIYqOxtYtkw+nj4dcHBQth4iIiIjxkBkqHbvBq5cARwdgSFDlK6GiIjIqDEQGarvvpP3w4cDtraKlkJERGTsGIgMkRBy1XoA6NlT2VqIiIhMAAORITp/HoiPB6ytgbZtla6GiIjI6DEQGaK9e+W9vz+gVitaChERkSlgIDJEe/bI+y5dlK2DiIjIRDAQGZrbt4GwMPm4Rw9layEiIjIRDESGZssWOQdR8+aAt7fS1RAREZkEBiJDExoq7wcOVLYOIiIiE2LQgWju3Llo2bIl7Ozs4OLigt69e+PixYt6bYQQmDVrFtzd3WFtbQ1/f39ERUXptUlPT8eECRNQuXJl2NraolevXrhx40ZZHkrh5OQAhw/Lx336KFsLERGRCTHoQBQeHo5x48bh2LFj2LdvH7KystClSxc8fPhQ12bBggVYuHAhli5dihMnTkCj0aBz585ITk7WtQkODsa2bdsQGhqKiIgIPHjwAIGBgcjOzlbisAqWmipDEQBoNMrWQkREZEJUQgihdBGFdfv2bbi4uCA8PBzt27eHEALu7u4IDg7Gu+++C0D2Brm6umL+/PkYNWoUtFotqlSpgo0bN2LAgAEAgNjYWHh4eGDXrl3o2rVroT47KSkJDg4O0Gq1sLe3L50DvHsXqFxZPs7KAszMSudziIiITERh/34bdA/R47RaLQDAyckJABAdHY34+Hh0+dfl6Wq1Gn5+fjhy5AgAIDIyEpmZmXpt3N3d4ePjo2uTn/T0dCQlJendSl1qqry3sGAYIiIiKkPlJhAJITBx4kS0a9cOPj4+AID4+HgAgKurq15bV1dX3XPx8fGwtLREpUqVCmyTn7lz58LBwUF38/DwKMnDyV9uILK2Lv3PIiIiIp1yE4jGjx+Ps2fP4ptvvsnznEql0tsWQuTZ97intZk2bRq0Wq3uFhMTU7zCi4KBiIiISBHlIhBNmDABP/30E8LCwlCtWjXdfs0/A48f7+lJSEjQ9RppNBpkZGQgMTGxwDb5UavVsLe317uVurQ0ec9AREREVKYMOhAJITB+/Hhs3boVBw8ehJeXl97zXl5e0Gg02Je7MjyAjIwMhIeHw9fXFwDQvHlzWFhY6LWJi4vD+fPndW0MBnuIiIiIFGGudAFPMm7cOGzevBnbt2+HnZ2drifIwcEB1tbWUKlUCA4ORkhICLy9veHt7Y2QkBDY2Nhg8ODBurYjR47EpEmT4OzsDCcnJ0yePBmNGjVCp06dlDy8vHIDkZWVsnUQERGZGIMORCtWrAAA+Pv76+1ft24dhg8fDgCYMmUKUlNTMXbsWCQmJqJ169bYu3cv7OzsdO0XLVoEc3Nz9O/fH6mpqQgICMD69ethZmhXcrGHiIiISBHlah4iJZXJPETffAMMHgwEBAD795fOZxAREZkQo5yHyOixh4iIiEgRDESGhGOIiIiIFMFAZEjYQ0RERKQIBiJDwnmIiIiIFMFAZEjYQ0RERKQIBiJDwjFEREREimAgMiTsISIiIlIEA5Eh4RgiIiIiRTAQGRL2EBERESmCgciQcAwRERGRIhiIDAl7iIiIiBTBQGRIOIaIiIhIEQxEhoQ9RERERIpgIDIkHENERESkCAYiQ8IeIiIiIkUwEBkSjiEiIiJSBAORIWEPERERkSIYiAwJxxAREREpgoHIUAjBU2ZEREQKYSAyFLlhCGAgIiIiKmMMRIaCgYiIiEgxDESGInf8kJkZYG6ubC1EREQmhoHIUPAKMyIiIsUwEBkKBiIiIiLFMBAZCl5hRkREpBgGIkPBOYiIiIgUw0BkKJKT5b2dnbJ1EBERmSAGIkORlCTv7e2VrYOIiMgEMRAZitweIgYiIiKiMsdAZChye4h4yoyIiKjMMRAZCp4yIyIiUgwDkaHgoGoiIiLFMBAZCvYQERERKYaByFBwUDUREZFiGIgMBQdVExERKYaByFDwlBkREZFiGIgMBQdVExERKYaByFCwh4iIiEgxDESGgoGIiIhIMQxEhkAInjIjIiJSEAORIUhLA7Ky5GP2EBEREZU5BiJDkNs7BAC2tsrVQUREZKIYiAzBv+cgqsCfhIiIqKzxr68h4IBqIiIiRTEQGQIOqCYiIlIUA5EhYA8RERGRohiIDAHXMSMiIlIUA5Eh4Er3REREimIgMgRarbxnDxEREZEiGIgMwalT8r5mTWXrICIiMlEMREoTAggPl4/9/RUthYiIyFSZVCBavnw5vLy8YGVlhebNm+Pw4cNKlwT8/TcQHw+o1UDr1kpXQ0REZJJMJhB9++23CA4OxnvvvYdTp07hhRdeQLdu3XD9+nVlC8vtHXr+ecDKStlaiIiITJTJBKKFCxdi5MiReP3111G/fn0sXrwYHh4eWLFihbKF/fqrvOfpMiIiIsWYRCDKyMhAZGQkunTpore/S5cuOHLkSL6vSU9PR1JSkt6tVDx4INcvYyAiIiJSjEkEojt37iA7Oxuurq56+11dXREfH5/va+bOnQsHBwfdzcPDo3SK++knIDERaNu2dN6fiIiInsokAlEulUqlty2EyLMv17Rp06DVanW3mJiY0ivM3h6wsCi99yciIqInMle6gLJQuXJlmJmZ5ekNSkhIyNNrlEutVkOtVpdFeURERKQwk+ghsrS0RPPmzbFv3z69/fv27YOvr69CVREREZGhMIkeIgCYOHEiXn31VbRo0QJt2rTBqlWrcP36dYwePVrp0oiIiEhhJhOIBgwYgLt37+KDDz5AXFwcfHx8sGvXLnh6eipdGhERESlMJYQQShdRHiQlJcHBwQFarRb2XJWeiIioXCjs32+TGENERERE9CQMRERERGTyGIiIiIjI5DEQERERkcljICIiIiKTx0BEREREJo+BiIiIiEweAxERERGZPJOZqfpZ5c5fmZSUpHAlREREVFi5f7efNg81A1EhJScnAwA8PDwUroSIiIiKKjk5GQ4ODgU+z6U7CiknJwexsbGws7ODSqUqsfdNSkqCh4cHYmJiTGpJEB43j9sU8Lh53KbA0I9bCIHk5GS4u7ujQoWCRwqxh6iQKlSogGrVqpXa+9vb2xvkP6TSxuM2LTxu08LjNi2GfNxP6hnKxUHVREREZPIYiIiIiMjkMRApTK1WY+bMmVCr1UqXUqZ43DxuU8Dj5nGbAmM5bg6qJiIiIpPHHiIiIiIyeQxEREREZPIYiIiIiMjkMRARERGRyWMgUtjy5cvh5eUFKysrNG/eHIcPH1a6pBIza9YsqFQqvZtGo9E9L4TArFmz4O7uDmtra/j7+yMqKkrBiovn0KFD6NmzJ9zd3aFSqfDjjz/qPV+Y40xPT8eECRNQuXJl2NraolevXrhx40YZHkXRPe24hw8fnuf3f/755/XalMfjnjt3Llq2bAk7Ozu4uLigd+/euHjxol4bY/zNC3Pcxvibr1ixAo0bN9ZNOtimTRvs3r1b97wx/tbA04/bGH9rBiIFffvttwgODsZ7772HU6dO4YUXXkC3bt1w/fp1pUsrMQ0bNkRcXJzudu7cOd1zCxYswMKFC7F06VKcOHECGo0GnTt31q0bV148fPgQTZo0wdKlS/N9vjDHGRwcjG3btiE0NBQRERF48OABAgMDkZ2dXVaHUWRPO24AePHFF/V+/127duk9Xx6POzw8HOPGjcOxY8ewb98+ZGVloUuXLnj48KGujTH+5oU5bsD4fvNq1aph3rx5OHnyJE6ePImOHTvipZde0oUeY/ytgacfN2B8vzUEKaZVq1Zi9OjRevvq1asnpk6dqlBFJWvmzJmiSZMm+T6Xk5MjNBqNmDdvnm5fWlqacHBwECtXriyjCkseALFt2zbddmGO8/79+8LCwkKEhobq2ty8eVNUqFBB7Nmzp8xqfxaPH7cQQgQFBYmXXnqpwNcYw3ELIURCQoIAIMLDw4UQpvObP37cQpjOb16pUiWxevVqk/mtc+UetxDG+Vuzh0ghGRkZiIyMRJcuXfT2d+nSBUeOHFGoqpJ36dIluLu7w8vLCwMHDsSVK1cAANHR0YiPj9c7frVaDT8/P6M6/sIcZ2RkJDIzM/XauLu7w8fHp9x/F7/++itcXFxQp04dvPHGG0hISNA9ZyzHrdVqAQBOTk4ATOc3f/y4cxnzb56dnY3Q0FA8fPgQbdq0MZnf+vHjzmVsvzUXd1XInTt3kJ2dDVdXV739rq6uiI+PV6iqktW6dWt89dVXqFOnDm7duoU5c+bA19cXUVFRumPM7/ivXbumRLmlojDHGR8fD0tLS1SqVClPm/L8b6Fbt2545ZVX4OnpiejoaMyYMQMdO3ZEZGQk1Gq1URy3EAITJ05Eu3bt4OPjA8A0fvP8jhsw3t/83LlzaNOmDdLS0lCxYkVs27YNDRo00P1hN9bfuqDjBozzt2YgUphKpdLbFkLk2VdedevWTfe4UaNGaNOmDWrVqoUNGzboBt8Z8/H/W3GOs7x/FwMGDNA99vHxQYsWLeDp6YmdO3eib9++Bb6uPB33+PHjcfbsWUREROR5zph/84KO21h/87p16+L06dO4f/8+tmzZgqCgIISHh+ueN9bfuqDjbtCggVH+1jxlppDKlSvDzMwsT1JOSEjI838bxsLW1haNGjXCpUuXdFebGfvxF+Y4NRoNMjIykJiYWGAbY+Dm5gZPT09cunQJQPk/7gkTJuCnn35CWFgYqlWrpttv7L95QcedH2P5zS0tLVG7dm20aNECc+fORZMmTfDZZ58Z/W9d0HHnxxh+awYihVhaWqJ58+bYt2+f3v59+/bB19dXoapKV3p6Ov7880+4ubnBy8sLGo1G7/gzMjIQHh5uVMdfmONs3rw5LCws9NrExcXh/PnzRvVd3L17FzExMXBzcwNQfo9bCIHx48dj69atOHjwILy8vPSeN9bf/GnHnR9j+c0fJ4RAenq60f7WBck97vwYxW9d5sO4SSc0NFRYWFiINWvWiAsXLojg4GBha2srrl69qnRpJWLSpEni119/FVeuXBHHjh0TgYGBws7OTnd88+bNEw4ODmLr1q3i3LlzYtCgQcLNzU0kJSUpXHnRJCcni1OnTolTp04JAGLhwoXi1KlT4tq1a0KIwh3n6NGjRbVq1cT+/fvFH3/8ITp27CiaNGkisrKylDqsp3rScScnJ4tJkyaJI0eOiOjoaBEWFibatGkjqlatWu6Pe8yYMcLBwUH8+uuvIi4uTndLSUnRtTHG3/xpx22sv/m0adPEoUOHRHR0tDh79qyYPn26qFChgti7d68Qwjh/ayGefNzG+lszECls2bJlwtPTU1haWornnntO7xLW8m7AgAHCzc1NWFhYCHd3d9G3b18RFRWlez4nJ0fMnDlTaDQaoVarRfv27cW5c+cUrLh4wsLCBIA8t6CgICFE4Y4zNTVVjB8/Xjg5OQlra2sRGBgorl+/rsDRFN6TjjslJUV06dJFVKlSRVhYWIjq1auLoKCgPMdUHo87v2MGINatW6drY4y/+dOO21h/8xEjRuj+G12lShUREBCgC0NCGOdvLcSTj9tYf2uVEEKUXX8UERERkeHhGCIiIiIyeQxEREREZPIYiIiIiMjkMRARERGRyWMgIiIiIpPHQEREREQmj4GIiIiITB4DEREREZk8BiIiMmn+/v4IDg5WugwiUhgDEREREZk8Lt1BRCZr+PDh2LBhg96+6Oho1KhRQ5mCiEgxDEREZLK0Wi26desGHx8ffPDBBwCAKlWqwMzMTOHKiKismStdABGRUhwcHGBpaQkbGxtoNBqlyyEiBXEMEREREZk8BiIiIiIyeQxERGTSLC0tkZ2drXQZRKQwBiIiMmk1atTA8ePHcfXqVdy5cwc5OTlKl0RECmAgIiKTNnnyZJiZmaFBgwaoUqUKrl+/rnRJRKQAXnZPREREJo89RERERGTyGIiIiIjI5DEQERERkcljICIiIiKTx0BEREREJo+BiIiIiEweAxERERGZPAYiIiIiMnkMRERERGTyGIiIiIjI5DEQERERkcljICIiIiKT93/neySK4JlbIgAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(0)\n",
    "plt.xlabel(\"t\")\n",
    "plt.ylabel(\"Regret\")\n",
    "plt.title(\"All classes with one Learner\")\n",
    "plt.plot(np.cumsum(opt_3_classes-np.mean(ts_rewards_per_experiment,axis=0), axis=0), 'r')\n",
    "plt.legend([\"TS\"])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 307,
   "metadata": {},
   "outputs": [],
   "source": [
    "#We have better performance when we consider the three classes one at time (because we use three learners instead of one)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "UCB"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "UCB with split2 (3 classes one at time, 3 learners)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 308,
   "metadata": {},
   "outputs": [],
   "source": [
    "#UCB class 0 + 1 + 2 for tot days\n",
    "\n",
    "E = E_commerce(binary_features=30)\n",
    "E.set_lambda(0.5)\n",
    "E.set_products(products)\n",
    "E.graph = np.array([[0. , 0.5, 0. , 1. , 0. ],[1. , 0. , 0. , 0.5, 0. ],[0. , 1. , 0. , 0.5, 0. ],[0. , 0.5, 1. , 0. , 0. ],[1. , 0. , 0. , 0.5, 0. ]])\n",
    "\n",
    "n_arms = 4\n",
    "num_users = 30*3\n",
    "n_days = 300\n",
    "fixed_alpha = 0\n",
    "fixed_weights = 1\n",
    "fixed_units = 0\n",
    "n_experiments = 10\n",
    "\n",
    "ucb3_rewards_per_experiment = []\n",
    "\n",
    "\n",
    "ucb3_pulls_per_arm_per_experiment = []\n",
    "\n",
    "\n",
    "\n",
    "for e in range(n_experiments):\n",
    "\n",
    "    env = Environment(n_arms, E, margins_matrix, num_users, fixed_alpha,fixed_weights,fixed_units)\n",
    "    ts_learner = UCB(n_arms=n_arms)\n",
    "\n",
    "    for d in range(n_days):\n",
    "        \n",
    "        pulled_arm = ts_learner.pull_arm()\n",
    "        reward = env.round(pulled_arm)\n",
    "        ts_learner.update(pulled_arm, reward)\n",
    "\n",
    "    ucb3_rewards_per_experiment.append(ts_learner.collected_rewards)\n",
    "\n",
    "    ucb3_pulls_per_arm_per_experiment.append(ts_learner.counter_per_arm)\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "E = E_commerce(binary_features=20)\n",
    "E.set_lambda(0.5)\n",
    "E.set_products(products)\n",
    "E.graph = np.array([[0. , 0.5, 0. , 1. , 0. ],[1. , 0. , 0. , 0.5, 0. ],[0. , 1. , 0. , 0.5, 0. ],[0. , 0.5, 1. , 0. , 0. ],[1. , 0. , 0. , 0.5, 0. ]])\n",
    "\n",
    "n_arms = 4\n",
    "num_users = 40*3\n",
    "n_days = 300\n",
    "fixed_alpha = 0\n",
    "fixed_weights = 1\n",
    "fixed_units = 0\n",
    "n_experiments = 10\n",
    "\n",
    "ucb2_rewards_per_experiment = []\n",
    "\n",
    "\n",
    "ucb2_pulls_per_arm_per_experiment = []\n",
    "\n",
    "\n",
    "\n",
    "for e in range(n_experiments):\n",
    "\n",
    "    env = Environment(n_arms, E, margins_matrix, num_users, fixed_alpha,fixed_weights,fixed_units)\n",
    "    ts_learner = UCB(n_arms=n_arms)\n",
    "\n",
    "    for d in range(n_days):\n",
    "        \n",
    "        pulled_arm = ts_learner.pull_arm()\n",
    "        reward = env.round(pulled_arm)\n",
    "        ts_learner.update(pulled_arm, reward)\n",
    "\n",
    "    ucb2_rewards_per_experiment.append(ts_learner.collected_rewards)\n",
    "\n",
    "    ucb2_pulls_per_arm_per_experiment.append(ts_learner.counter_per_arm)\n",
    "\n",
    "\n",
    "\n",
    "E = E_commerce(binary_features=10)\n",
    "E.set_lambda(0.5)\n",
    "E.set_products(products)\n",
    "E.graph = np.array([[0. , 0.5, 0. , 1. , 0. ],[1. , 0. , 0. , 0.5, 0. ],[0. , 1. , 0. , 0.5, 0. ],[0. , 0.5, 1. , 0. , 0. ],[1. , 0. , 0. , 0.5, 0. ]])\n",
    "\n",
    "n_arms = 4\n",
    "num_users = 30*3\n",
    "n_days = 300\n",
    "fixed_alpha = 0\n",
    "fixed_weights = 1\n",
    "fixed_units = 0\n",
    "n_experiments = 10\n",
    "\n",
    "ucb1_rewards_per_experiment = []\n",
    "\n",
    "\n",
    "ucb1_pulls_per_arm_per_experiment = []\n",
    "\n",
    "\n",
    "for e in range(n_experiments):\n",
    "\n",
    "    env = Environment(n_arms, E, margins_matrix, num_users, fixed_alpha,fixed_weights,fixed_units)\n",
    "    ts_learner = UCB(n_arms=n_arms)\n",
    "\n",
    "    for d in range(n_days):\n",
    "        \n",
    "        pulled_arm = ts_learner.pull_arm()\n",
    "        reward = env.round(pulled_arm)\n",
    "        ts_learner.update(pulled_arm, reward)\n",
    "\n",
    "    ucb1_rewards_per_experiment.append(ts_learner.collected_rewards)\n",
    "\n",
    "    ucb1_pulls_per_arm_per_experiment.append(ts_learner.counter_per_arm)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 309,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TS 141.43407433789957 14.13719457919283\n",
      "TS 120.46212729223743 12.205016210565221\n",
      "TS 107.24539356621005 11.351312159602294\n"
     ]
    }
   ],
   "source": [
    "print(\"TS\",np.mean(np.mean(ts1_rewards_per_experiment, axis=0)), np.std(np.mean(ts1_rewards_per_experiment, axis=0)))\n",
    "print(\"TS\",np.mean(np.mean(ts2_rewards_per_experiment, axis=0)), np.std(np.mean(ts2_rewards_per_experiment, axis=0)))\n",
    "print(\"TS\",np.mean(np.mean(ts3_rewards_per_experiment, axis=0)), np.std(np.mean(ts3_rewards_per_experiment, axis=0)))\n",
    "\n",
    "ucb_all_classes_rewards = 0.3*(np.mean(ucb1_rewards_per_experiment, axis=0))+0.4*(np.mean(ucb2_rewards_per_experiment, axis=0))+0.3*np.mean(ucb3_rewards_per_experiment, axis=0)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 310,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(0)\n",
    "plt.xlabel(\"t\")\n",
    "plt.ylabel(\"Regret\")\n",
    "plt.plot(np.cumsum(opt123-ucb_all_classes_rewards, axis=0), 'b')\n",
    "plt.legend([\"UCB\"])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 311,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_______________________________________________\n",
      "Considering all the experiments:\n",
      "-----------------------------------------------\n",
      "Total expected reward over 300 days\n",
      "36506.887817166666\n",
      "-----------------------------------------------\n",
      "Total expected regret over 300 days\n",
      "1193.552432194375\n",
      "-----------------------------------------------\n",
      "Average expected reward per round over 300 days\n",
      "121.68962605722221\n",
      "-----------------------------------------------\n",
      "Average expected regret per round over 300 days\n",
      "3.978508107314583\n"
     ]
    }
   ],
   "source": [
    "print(\"_______________________________________________\")\n",
    "print(\"Considering all the experiments:\")\n",
    "print(\"-----------------------------------------------\")\n",
    "print(\"Total expected reward over\",n_days, \"days\")\n",
    "print(np.mean(np.sum(ucb_all_classes_rewards)))\n",
    "print(\"-----------------------------------------------\")\n",
    "print(\"Total expected regret over\",n_days, \"days\")\n",
    "print(np.mean(np.sum(opt123 - ucb_all_classes_rewards)))\n",
    "\n",
    "print(\"-----------------------------------------------\")\n",
    "print(\"Average expected reward per round over\",n_days, \"days\")\n",
    "print(np.mean(np.mean(ucb_all_classes_rewards, axis=0)))\n",
    "print(\"-----------------------------------------------\")\n",
    "print(\"Average expected regret per round over\",n_days, \"days\")\n",
    "print(np.mean(np.mean(opt123-ucb_all_classes_rewards, axis=0)))\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.9.7 ('base')",
   "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": "3.9.7"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "f1838362fc3c72254db1311c9d5db56b79b6520b93de3800103ebfd8112b592e"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}