cancer_ope.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "from tqdm.notebook import tqdm\n",
    "import pickle\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.neighbors import NearestNeighbors"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from src.envs.cancer import CancerSim, PolicyCancer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# with open(\"data/cancer_pcv_train_episodes\", 'rb') as f:\n",
    "#     train_dataset = pickle.load(f)\n",
    "# with open(\"data/cancer_pcv_val_episodes\", 'rb') as f:\n",
    "#     val_dataset = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "from src.offline_ql import discrete_PQL\n",
    "from src.offline_pg import ISPG"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def online_evaluate(env, policy, trees, sample_size=1000):\n",
    "    returns = []\n",
    "    action_dict = dict()\n",
    "    avg_treatment = []\n",
    "    change_in_mtd = []\n",
    "    for i in range(sample_size):\n",
    "        obs = env.reset()\n",
    "        done = False\n",
    "        rt = 0\n",
    "        t = 0\n",
    "        treatment = []\n",
    "        init_mtd = obs[1]\n",
    "        while not done:\n",
    "            nn_action_dist = get_nn_action_dist(obs, trees)\n",
    "            pibs = PolicyCancer().return_probs(obs, t)\n",
    "            a = policy.select_action(obs, pibs, nn_action_dist)\n",
    "            t += 1\n",
    "            obs, reward, done, _ = env.step(a)\n",
    "            rt += reward\n",
    "            treatment.append(a)\n",
    "        end_mtd = obs[1]\n",
    "        returns.append(rt)\n",
    "        avg_treatment.append(treatment)\n",
    "        change_in_mtd.append(100*(end_mtd-init_mtd)/init_mtd)\n",
    "    avg_treatment = np.array(avg_treatment).mean(axis=0)\n",
    "    avg_change_in_mtd = np.mean(change_in_mtd)\n",
    "    return np.mean(returns), np.std(returns), avg_treatment, avg_change_in_mtd\n",
    "\n",
    "\n",
    "def get_nn_action_dist(obs, trees):\n",
    "    nn_action_dist = np.zeros(len(trees))\n",
    "    for a in range(len(trees)):\n",
    "        nn_action_dist[a] = (trees[a].kneighbors(obs.reshape(1,-1))[0].item()**2)/obs.size\n",
    "    return nn_action_dist\n",
    "\n",
    "\n",
    "def construct_nearest_neighbor_trees(observations, actions, num_actions=2):\n",
    "    trees = dict()\n",
    "    for a in range(num_actions):\n",
    "        trees[a] = NearestNeighbors(n_neighbors=1)\n",
    "        is_action = (np.squeeze(actions,axis=-1) == a)\n",
    "        trees[a].fit(observations[is_action, :])\n",
    "    return trees\n",
    "\n",
    "def test_fix_policy(env, eps=0, n=1000):\n",
    "    policy = PolicyCancer(eps_behavior=eps)\n",
    "    rewards = []\n",
    "    for i in range(n):\n",
    "        obs = env.reset()\n",
    "        t = 0\n",
    "        reward = 0\n",
    "        done = False\n",
    "        while not done:\n",
    "            a = policy(obs, t)\n",
    "            obs, rt, done, _ = env.step(a)\n",
    "            t += 1\n",
    "            reward += rt\n",
    "        rewards.append(reward)\n",
    "    return np.mean(rewards), np.std(rewards)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def wis_fix_policy(dataset):\n",
    "    policy = PolicyCancer(eps_behavior=0.01)\n",
    "\n",
    "    n = dataset[\"observations\"].shape[0]\n",
    "    horizon = dataset[\"observations\"].shape[1]\n",
    "    weights = np.ones((n, horizon))\n",
    "    \n",
    "    rewards = dataset[\"rewards\"][:,:,0]\n",
    "    \n",
    "    for idx in range(n):\n",
    "        last = 1\n",
    "        for t in range(horizon):\n",
    "            pie = policy.return_probs(dataset[\"observations\"][idx,t,:], t)\n",
    "            a = dataset[\"actions\"][idx,t,0]\n",
    "            \n",
    "            weights[idx, t] = last * (pie[a] / dataset[\"pibs\"][idx, t, a])\n",
    "            last = weights[idx, t]\n",
    "\n",
    "            if not dataset[\"not_done\"][idx, t, 0]:\n",
    "                weights[idx, t+1:] = weights[idx, t]\n",
    "                break\n",
    "    \n",
    "    weights = np.clip(weights, 0, 1e3)\n",
    "    weights_norm = weights.sum(axis=0)\n",
    "    weights /= weights_norm\n",
    "    estm = (weights[:,-1] * rewards.sum(axis=-1)).sum()\n",
    "    \n",
    "    return estm\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_model_from_param(param):\n",
    "    \n",
    "    FEATURES_UB = [3.25401319, 51.84731224, 30.]\n",
    "    # FEATURES_UB = [3.18970173,  7.85591301, 44.83595683, 30.64735325, 30.]\n",
    "    \n",
    "    if param[\"name\"] == \"PG_mask\":\n",
    "        settings = f\"PG_{param['env']}_masknn_action_dist_weightedTrueam0.0\"\\\n",
    "                   f\"thresh{param['thresh']}var{param['var_coeff']}clip0.01seed{param['seed']}\"\n",
    "        model = ISPG(state_dim=3, num_actions=2, device=\"cpu\", hidden_dim=32,\n",
    "                     threshold=param['thresh'], action_mask_type=\"nn_action_dist\")\n",
    "        model.load(f\"models/{settings}_best_\")\n",
    "        val_score = np.load(f\"./results/{settings}_val_wis_scores.npy\")\n",
    "        train_score = np.load(f\"./results/{settings}_train_wis_scores.npy\")[:,0]\n",
    "        test_score = np.load(f\"./results/{settings}_true_scores.npy\") \n",
    "    \n",
    "    elif param[\"name\"] == \"PG_nomask\":\n",
    "        settings = f\"PG_{param['env']}_maskstep_weightedTrueam0.0\"\\\n",
    "                   f\"thresh{param['thresh']}var{param['var_coeff']}clip0.01seed{param['seed']}\"\n",
    "        model = ISPG(state_dim=3, num_actions=2, device=\"cpu\", hidden_dim=32,\n",
    "                     threshold=param['thresh'], action_mask_type=\"step\")\n",
    "        model.load(f\"models/{settings}_best_\")\n",
    "        val_score = np.load(f\"./results/{settings}_val_wis_scores.npy\")\n",
    "        train_score = np.load(f\"./results/{settings}_train_wis_scores.npy\")[:,0]\n",
    "        test_score = np.load(f\"./results/{settings}_true_scores.npy\") \n",
    "    \n",
    "    elif param[\"name\"] == \"PG_pib_mask\":\n",
    "        model = ISPG(state_dim=3, num_actions=2, device=\"cpu\", hidden_dim=32,\n",
    "                     threshold=param['thresh'], action_mask_type=\"step\")\n",
    "        val_score = None\n",
    "        train_score = None\n",
    "        test_score = None\n",
    "    \n",
    "    elif param[\"name\"] == \"BCQ\":\n",
    "        settings = f\"BCQ_{param['env']}_thresh{param['thresh']}_seed{param['seed']}\"\n",
    "        model = discrete_PQL(state_dim=3, num_actions=2, device=\"cpu\", state_clipping=False, hidden_dim=32,\n",
    "                             max_state=FEATURES_UB, action_threshold=param['thresh'], log_pibs=True)\n",
    "        model.load(f\"models/{settings}_best_\")\n",
    "        val_score = np.load(f\"./results/{settings}_val_scores.npy\")\n",
    "        train_score = np.load(f\"./results/{settings}_q_scores.npy\")\n",
    "        test_score = np.load(f\"./results/{settings}_true_scores.npy\") \n",
    "    \n",
    "    elif param[\"name\"] == \"PQL\":\n",
    "        settings = f\"PQL_{param['env']}_thresh{param['thresh']}_seed{param['seed']}\"\n",
    "        model = discrete_PQL(state_dim=3, num_actions=2, device=\"cpu\", state_clipping=True, hidden_dim=32,\n",
    "                             max_state=FEATURES_UB, action_threshold=param['thresh'], log_pibs=True)\n",
    "        model.load(f\"models/{settings}_best_\")\n",
    "        val_score = np.load(f\"./results/{settings}_val_scores.npy\")\n",
    "        train_score = np.load(f\"./results/{settings}_q_scores.npy\")\n",
    "        test_score = np.load(f\"./results/{settings}_true_scores.npy\") \n",
    "    \n",
    "    return model, val_score, train_score, test_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_model_from_param(param):\n",
    "    \n",
    "    #FEATURES_UB = [3.25401319, 51.84731224, 30.]\n",
    "    state_dim = 5\n",
    "    FEATURES_UB = [3.18970173,  7.85591301, 44.83595683, 30.64735325, 30.]\n",
    "    \n",
    "    if param[\"name\"] == \"PG_mask\":\n",
    "        settings = f\"PG_{param['env']}_masknn_action_dist_\"\\\n",
    "                   f\"thresh{param['thresh']}var{param['var_coeff']}seed{param['seed']}\"\n",
    "        model = ISPG(state_dim=state_dim, num_actions=2, device=\"cpu\", hidden_dim=32,\n",
    "                     threshold=param['thresh'], action_mask_type=\"nn_action_dist\")\n",
    "        model.load(f\"models/{settings}_best_\")\n",
    "        val_score = np.load(f\"./results/{settings}_val_wis_scores.npy\")\n",
    "        train_score = np.load(f\"./results/{settings}_train_wis_scores.npy\")[:,0]\n",
    "    \n",
    "    elif param[\"name\"] == \"PG_nomask\":\n",
    "        settings = f\"PG_{param['env']}_maskstep_\"\\\n",
    "                   f\"thresh{param['thresh']}var{param['var_coeff']}seed{param['seed']}\"\n",
    "        model = ISPG(state_dim=state_dim, num_actions=2, device=\"cpu\", hidden_dim=32,\n",
    "                     threshold=param['thresh'], action_mask_type=\"step\")\n",
    "        model.load(f\"models/{settings}_best_\")\n",
    "        val_score = np.load(f\"./results/{settings}_val_wis_scores.npy\")\n",
    "        train_score = np.load(f\"./results/{settings}_train_wis_scores.npy\")[:,0]\n",
    "    \n",
    "    elif param[\"name\"] == \"PG_pib_mask\":\n",
    "        model = ISPG(state_dim=state_dim, num_actions=2, device=\"cpu\", hidden_dim=32,\n",
    "                     threshold=param['thresh'], action_mask_type=\"step\")\n",
    "        val_score = None\n",
    "        train_score = None\n",
    "        test_score = None\n",
    "    \n",
    "    elif param[\"name\"] == \"BCQ\":\n",
    "        settings = f\"BCQ_{param['env']}_thresh{param['thresh']}_beta2.0_seed{param['seed']}\"\n",
    "        model = discrete_PQL(state_dim=state_dim, num_actions=2, device=\"cpu\", state_clipping=False, hidden_dim=32,\n",
    "                             max_state=FEATURES_UB, action_threshold=param['thresh'], log_pibs=True)\n",
    "        model.load(f\"models/{settings}_best_\")\n",
    "        val_score = np.load(f\"./results/{settings}_val_scores.npy\")\n",
    "        train_score = np.expand_dims(np.load(f\"./results/{settings}_q_scores.npy\"), axis=1)\n",
    "        #val_score = val_score[:5,:]\n",
    "        #val_score = np.hstack((train_score, val_score))\n",
    "    \n",
    "    elif param[\"name\"] == \"PQL\":\n",
    "        settings = f\"PQL_{param['env']}_thresh{param['thresh']}_beta{param['beta']}_seed{param['seed']}\"\n",
    "        model = discrete_PQL(state_dim=state_dim, num_actions=2, device=\"cpu\", state_clipping=True, hidden_dim=32,\n",
    "                             max_state=FEATURES_UB, action_threshold=param['thresh'], log_pibs=True)\n",
    "        model.load(f\"models/{settings}_best_\")\n",
    "        val_score = np.load(f\"./results/{settings}_val_scores.npy\")\n",
    "        train_score = np.expand_dims(np.load(f\"./results/{settings}_q_scores.npy\"),axis=1)\n",
    "        #val_score = val_score[:5,:]\n",
    "        #val_score = np.hstack((train_score, val_score))\n",
    "    \n",
    "    test_score = np.load(f\"./results/{settings}_true_scores.npy\")\n",
    "    return model, val_score, train_score, test_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_best_hp(env, alg, verbose=True):\n",
    "    #var_coeff = 0.0\n",
    "    var_coeff_list = [0.0, 0.1, 1.0]\n",
    "    seeds = [1,2,3,4,5]\n",
    "    if alg == \"PG_mask\":\n",
    "        threshold_list = [0.05, 0.1, 0.5]\n",
    "    elif alg == \"PG_nomask\":\n",
    "        threshold_list = [0.0]\n",
    "    elif alg == \"PG_pib_mask\":\n",
    "        with open(f\"data/{env_name}_train_episodes\", 'rb') as f:\n",
    "            train_dataset = pickle.load(f)\n",
    "        with open(f\"data/{env_name}_val_episodes\", 'rb') as f:\n",
    "            val_dataset = pickle.load(f)\n",
    "        return {\"name\": alg, \"env\": env, \"thresh\": 0.2, \"var_coeff\": 0, \"seed\": 0}, wis_fix_policy(val_dataset), wis_fix_policy(val_dataset)\n",
    "    elif alg == \"BCQ\":\n",
    "        var_coeff_list = [0.0]\n",
    "        threshold_list = [0.0, 0.25]\n",
    "    elif alg == \"PQL\":\n",
    "        var_coeff_list = [1.0, 2.0, 5.0]\n",
    "        threshold_list = [0.0, 0.25]\n",
    "    best_wis = -100\n",
    "    for var_coeff in var_coeff_list:\n",
    "        for threshold in threshold_list:\n",
    "            for seed in seeds:\n",
    "                if alg == \"PQL\":\n",
    "                    param = {\"name\": alg, \"env\": env, \"thresh\": threshold, \"beta\": var_coeff, \"seed\": seed}\n",
    "                else:\n",
    "                    param = {\"name\": alg, \"env\": env, \"thresh\": threshold, \"var_coeff\": var_coeff, \"seed\": seed}\n",
    "                model, score, train_score, _ = get_model_from_param(param)\n",
    "        \n",
    "                wis_pie = score[:,0].max()\n",
    "                ESS = score[np.argmax(score[:,0]),-1]\n",
    "                train = train_score[np.argmax(score[:,0])]\n",
    "            \n",
    "                if wis_pie > best_wis:\n",
    "                    best_wis = wis_pie\n",
    "                    best_train = train\n",
    "                    best_param = param\n",
    "                    if verbose:\n",
    "                        print(\"New wis:\", round(wis_pie, 2), \"ESS:\", round(ESS, 2), \"at:\", param)\n",
    "    return best_param, best_wis, best_train\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "New wis: 2.2 ESS: 2.01 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv1', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 2.34 ESS: 1.89 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv1', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 3}\n",
      "New wis: 2.64 ESS: 1.34 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv1', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 5}\n",
      "New wis: 2.96 ESS: 1.26 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv1', 'thresh': 0.5, 'var_coeff': 0.0, 'seed': 5}\n",
      "Online Test: cancer_mdp_pcv1 PG_mask 0.13412338095166182\n",
      "New wis: 2.32 ESS: 2.31 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 3.0 ESS: 1.18 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 5}\n",
      "New wis: 3.01 ESS: 1.16 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'var_coeff': 0.1, 'seed': 5}\n",
      "New wis: 3.01 ESS: 1.15 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'var_coeff': 1.0, 'seed': 5}\n",
      "Online Test: cancer_mdp_pcv1 PG_nomask -1.3713050410116912\n",
      "Online Test: cancer_mdp_pcv1 PG_pib_mask -4.350138017843129\n",
      "New wis: 3.06 ESS: 1.01 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 3.07 ESS: 1.02 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 2}\n",
      "New wis: 3.08 ESS: 1.01 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 3}\n",
      "New wis: 3.08 ESS: 1.0 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 5}\n",
      "Online Test: cancer_mdp_pcv1 BCQ 2.9557533260653055\n",
      "New wis: 3.07 ESS: 1.01 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'beta': 1.0, 'seed': 1}\n",
      "New wis: 3.07 ESS: 1.02 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'beta': 1.0, 'seed': 2}\n",
      "New wis: 3.08 ESS: 1.01 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'beta': 1.0, 'seed': 3}\n",
      "New wis: 3.08 ESS: 1.0 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'beta': 1.0, 'seed': 4}\n",
      "Online Test: cancer_mdp_pcv1 PQL 3.086715443366803\n",
      "New wis: -1.43 ESS: 14.28 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv2', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: -1.41 ESS: 4.6 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv2', 'thresh': 0.1, 'var_coeff': 0.0, 'seed': 5}\n",
      "New wis: 0.6 ESS: 1.61 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv2', 'thresh': 0.5, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 0.74 ESS: 2.11 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv2', 'thresh': 0.5, 'var_coeff': 0.0, 'seed': 5}\n",
      "New wis: 0.74 ESS: 2.1 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv2', 'thresh': 0.5, 'var_coeff': 0.1, 'seed': 5}\n",
      "New wis: 0.76 ESS: 2.06 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv2', 'thresh': 0.5, 'var_coeff': 1.0, 'seed': 5}\n",
      "Online Test: cancer_mdp_pcv2 PG_mask 0.7032732874484754\n",
      "New wis: 0.6 ESS: 1.58 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv2', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 0.76 ESS: 2.18 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv2', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 5}\n",
      "Online Test: cancer_mdp_pcv2 PG_nomask 0.10102065015408059\n",
      "Online Test: cancer_mdp_pcv2 PG_pib_mask -4.266489466438258\n",
      "New wis: 0.58 ESS: 1.0 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv2', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 1.84 ESS: 1.0 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv2', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 2}\n",
      "Online Test: cancer_mdp_pcv2 BCQ -3.4313981416498747\n",
      "New wis: 2.35 ESS: 1.02 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv2', 'thresh': 0.0, 'beta': 1.0, 'seed': 1}\n",
      "New wis: 2.35 ESS: 1.02 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv2', 'thresh': 0.0, 'beta': 2.0, 'seed': 1}\n",
      "New wis: 2.35 ESS: 1.02 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv2', 'thresh': 0.0, 'beta': 5.0, 'seed': 1}\n",
      "Online Test: cancer_mdp_pcv2 PQL 2.180178050469395\n",
      "New wis: 1.49 ESS: 2.45 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv3', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 1}\n",
      "Online Test: cancer_mdp_pcv3 PG_mask 0.7938896659001362\n",
      "New wis: 0.28 ESS: 3.72 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv3', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 0.96 ESS: 3.11 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv3', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 5}\n",
      "Online Test: cancer_mdp_pcv3 PG_nomask 1.4249453134457823\n",
      "Online Test: cancer_mdp_pcv3 PG_pib_mask -4.31964423499903\n",
      "New wis: 2.54 ESS: 1.12 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv3', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 2.63 ESS: 1.11 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv3', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 3}\n",
      "Online Test: cancer_mdp_pcv3 BCQ 0.5418232041767683\n",
      "New wis: 2.5 ESS: 1.21 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv3', 'thresh': 0.0, 'beta': 1.0, 'seed': 1}\n",
      "New wis: 2.5 ESS: 1.21 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv3', 'thresh': 0.0, 'beta': 2.0, 'seed': 1}\n",
      "New wis: 2.58 ESS: 1.11 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv3', 'thresh': 0.0, 'beta': 5.0, 'seed': 2}\n",
      "New wis: 2.65 ESS: 1.05 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv3', 'thresh': 0.0, 'beta': 5.0, 'seed': 4}\n",
      "Online Test: cancer_mdp_pcv3 PQL 0.6921276486952802\n",
      "New wis: -1.08 ESS: 5.87 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv4', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: -0.77 ESS: 3.04 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv4', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 5}\n",
      "New wis: -0.12 ESS: 5.53 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv4', 'thresh': 0.1, 'var_coeff': 0.0, 'seed': 1}\n",
      "Online Test: cancer_mdp_pcv4 PG_mask -1.09391321286853\n",
      "New wis: -0.48 ESS: 9.18 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: -0.35 ESS: 4.45 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 5}\n",
      "New wis: -0.35 ESS: 4.69 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'var_coeff': 0.1, 'seed': 5}\n",
      "Online Test: cancer_mdp_pcv4 PG_nomask -8.706243049752139\n",
      "Online Test: cancer_mdp_pcv4 PG_pib_mask -4.4579588687932565\n",
      "New wis: 0.31 ESS: 2.19 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 0.42 ESS: 1.39 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 2}\n",
      "New wis: 0.53 ESS: 1.05 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 3}\n",
      "Online Test: cancer_mdp_pcv4 BCQ 2.8078820136562213\n",
      "New wis: 0.19 ESS: 2.29 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'beta': 1.0, 'seed': 1}\n",
      "New wis: 0.41 ESS: 1.77 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'beta': 1.0, 'seed': 2}\n",
      "New wis: 0.42 ESS: 1.13 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'beta': 1.0, 'seed': 3}\n",
      "New wis: 0.61 ESS: 1.0 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'beta': 1.0, 'seed': 5}\n",
      "Online Test: cancer_mdp_pcv4 PQL 3.022404041939303\n",
      "New wis: -2.19 ESS: 6.86 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv5', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: -1.62 ESS: 1.96 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv5', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 4}\n",
      "New wis: 1.43 ESS: 1.12 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv5', 'thresh': 0.1, 'var_coeff': 0.0, 'seed': 1}\n",
      "Online Test: cancer_mdp_pcv5 PG_mask 0.09312748917040181\n",
      "New wis: 0.8 ESS: 1.73 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 1.01 ESS: 1.08 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 5}\n",
      "New wis: 1.01 ESS: 1.07 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'var_coeff': 0.1, 'seed': 5}\n",
      "New wis: 1.05 ESS: 1.03 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'var_coeff': 1.0, 'seed': 5}\n",
      "Online Test: cancer_mdp_pcv5 PG_nomask -3.2839752645226223\n",
      "Online Test: cancer_mdp_pcv5 PG_pib_mask -4.290517772778546\n",
      "New wis: 1.07 ESS: 1.98 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 1.18 ESS: 1.43 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 3}\n",
      "New wis: 1.52 ESS: 1.01 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 5}\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Online Test: cancer_mdp_pcv5 BCQ 2.4223678042389207\n",
      "New wis: 0.75 ESS: 1.08 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'beta': 1.0, 'seed': 1}\n",
      "New wis: 1.52 ESS: 1.01 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'beta': 1.0, 'seed': 2}\n",
      "New wis: 1.52 ESS: 1.01 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'beta': 2.0, 'seed': 2}\n",
      "Online Test: cancer_mdp_pcv5 PQL 2.9694913512342267\n"
     ]
    }
   ],
   "source": [
    "env_list = [\"cancer_mdp_pcv1\", \"cancer_mdp_pcv2\", \"cancer_mdp_pcv3\", \"cancer_mdp_pcv4\", \"cancer_mdp_pcv5\"]\n",
    "env = CancerSim(dose_penalty=1.0, therapy_type=\"PCV\", env_seed=None, max_steps=30, state_dim=5)\n",
    "\n",
    "#env_list = [\"cancer_pcv1\", \"cancer_pcv2\", \"cancer_pcv3\", \"cancer_pcv4\", \"cancer_pcv5\"]\n",
    "#env = CancerSim(dose_penalty=1.0, therapy_type=\"PCV\", env_seed=None, max_steps=30, state_dim=3)\n",
    "\n",
    "algo_list = [\"PG_mask\", \"PG_nomask\", \"PG_pib_mask\", \"BCQ\", \"PQL\"]\n",
    "\n",
    "test_results = np.zeros((len(algo_list), len(env_list)))\n",
    "train_results = np.zeros((len(algo_list), len(env_list)))\n",
    "valid_results = np.zeros((len(algo_list), len(env_list)))\n",
    "treatment_logs = np.zeros((len(algo_list), len(env_list), 30))\n",
    "mtd_changes = np.zeros((len(algo_list), len(env_list)))\n",
    "\n",
    "\n",
    "\n",
    "for j,env_name in enumerate(env_list):\n",
    "    with open(f\"data/{env_name}_train_episodes\", 'rb') as f:\n",
    "        train_dataset = pickle.load(f)\n",
    "    trees = construct_nearest_neighbor_trees(train_dataset[\"observations\"],train_dataset[\"actions\"])\n",
    "    for i,algo in enumerate(algo_list):\n",
    "        param, val, train = get_best_hp(env_name, algo)\n",
    "        model, _, _, _ = get_model_from_param(param)\n",
    "        test, _, treatment_log, mtd_change = online_evaluate(env, model, trees, sample_size=100)\n",
    "        print(\"Online Test:\", env_name, algo, test)\n",
    "        \n",
    "        test_results[i,j] = test\n",
    "        train_results[i,j] = train\n",
    "        valid_results[i,j] = val\n",
    "        treatment_logs[i,j,:] = treatment_log\n",
    "        mtd_changes[i,j] = mtd_change"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {},
   "outputs": [],
   "source": [
    "v_max = 2.5\n",
    "v_min = -20.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [],
   "source": [
    "final_test_results = 100*(test_results-v_min)/(v_max-v_min)\n",
    "final_valid_results = 100*(valid_results-v_min)/(v_max-v_min)\n",
    "final_train_results = 100*(train_results-v_min)/(v_max-v_min)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.set_printoptions(precision=2, linewidth=100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([89.52, 78.79, 69.46, 95.64, 96.88])"
      ]
     },
     "execution_count": 115,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "final_test_results[:,:].mean(axis=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1.55, 6.42, 0.26, 4.64, 3.76])"
      ]
     },
     "execution_count": 116,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "final_test_results.std(axis=-1)/np.sqrt(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 5.16, 14.93, -0.69,  0.55, -0.26])"
      ]
     },
     "execution_count": 117,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "final_valid_results.mean(axis=-1) - final_test_results.mean(axis=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1.78, 5.71, 0.81, 5.12, 4.18])"
      ]
     },
     "execution_count": 118,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(final_valid_results - final_test_results).std(axis=-1)/np.sqrt(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([7.61, 3.23, 9.  , 0.4 ])"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "treatment_logs[:,:,:].sum(axis=-1).mean(axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1.69, 0.49, 0.  , 0.36])"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "treatment_logs[:,:,:].sum(axis=-1).std(axis=1)/np.sqrt(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([-53.43,   6.44, -64.46,  89.81])"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mtd_changes.mean(axis=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 9.95, 12.99,  0.11, 10.72])"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mtd_changes.std(axis=-1)/np.sqrt(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXUAAAD4CAYAAAATpHZ6AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8li6FKAAAgAElEQVR4nOydd3iUVdr/P2d6Sy+UTAIEAglFQCIIKoJlQbGyNnDtP3ldwV1dd13ftexa1vrurr6La1nX1wp2kLWhC4oiKkVCTeglgZBO+mTa+f3xJCGBlCHMZDLhfK5rrmTmOfM8dwL5zv3c5z7fI6SUKBQKhaJ3oAt3AAqFQqEIHkrUFQqFohehRF2hUCh6EUrUFQqFohehRF2hUCh6EYZwXTgxMVEOHDgwXJdXKBSKiGTdunWlUsqk9o6HTdQHDhzI2rVrw3V5hUKhiEiEEPs6Oq7KLwqFQtGLUKKuUCgUvQgl6gqFQtGLCFtNXaFQ9Bw8Hg8FBQW4XK5wh6JoxGKx4HQ6MRqNx/U+JeoKhYKCggKioqIYOHAgQohwh3PSI6WkrKyMgoICBg0adFzv7bT8IoR4RQhRLITY3M5xIYT4XyHETiHERiHEqccVgUKhCDsul4uEhAQl6D0EIQQJCQldunMKpKb+KjC9g+MXABmNjznA88cdhUKhCDtK0HsWXf336LT8IqX8RggxsIMhlwKvS83D9wchRKwQop+UsrBLEXXCuy++Qen6g8RYOq8zSQH1Z0gaBptDEUqvYUD0AC4efHG4w1AoFEEgGDX1FCC/xfOCxteOEXUhxBy0bJ60tLQuXax8cyFSjuNwfQCDhQ7fZ6t5efJrXbrWyYBEIhBMHzQdo+74JmQUCkXPIxii3tY9Qps7b0gpXwJeAsjOzu7S7hxZd9zEDa+s5u05p3N6ekKHYz+88UUqzCPZcP0GdWvZDm9ufZMn1zxJnaeOGHNMuMNRnMTo9XpGjRqF1+slKyuL1157DZvNRlFREXfddRc//PADcXFxmEwm7rnnHi6//PJui+3VV19l7dq1zJ8/v9uu2VWC0adeAKS2eO4EDgbhvG2SnmgHYHdJbadj+zhqcQkbFYfqQhVOxGM3ar/PGk9NmCNRnOxYrVZycnLYvHkzJpOJF154ASkll112GZMnT2b37t2sW7eOt99+m4KCgnCH22MJRqa+BJgnhHgbmABUhqqeDpASa8Vs0LG7pHMR6p9iIGcf5G8tI76fPVQhRTTNou5Woq7QeOjfW9h6sCqo5xzeP5o/Xjwi4PFnnXUWGzduZPny5ZhMJm677bbmYwMGDOCOO+5o972vvvoqixcvxufzsXnzZu6++27cbjdvvPEGZrOZTz/9lPj4eP75z3/y0ksv4Xa7GTJkCG+88QY2m4333nuPhx56CL1eT0xMDN98802r83/yySc8+uij/Pvf/yYxMfH4fxkhJpCWxoXA98AwIUSBEOIWIcRtQoim3/KnwG5gJ/BP4PaQRQvodIJBiXZ2BSDqsYP6YK0vIX/joVCGFNE4jA4A6rzqbkbRM/B6vXz22WeMGjWKLVu2cOqpx98lvXnzZhYsWMDq1au57777sNlsrF+/nokTJ/L6668DMHPmTNasWcOGDRvIysriX//6FwAPP/wwS5cuZcOGDSxZsqTVeRctWsQTTzzBp59+2iMFHQLrfpnVyXEJzA1aRAEwOMnB5oOVnY4zpjqJq/gPB3cn4ff50emVK8LR2E0qU1e05ngy6mBSX1/PmDFjAC1Tv+WWW3jhhRdajZk7dy4rV67EZDKxZs2ads81depUoqKiiIqKIiYmhosv1rq7Ro0axcaNGwFN+O+//34OHz5MTU0N06ZNA+CMM87gxhtv5KqrrmLmzJnN5/zqq69Yu3YtX3zxBdHR0UH92YNJRKrc4CQ7+eV1NHh9HY4zpaYSX7ENjweK91V3U3SRhd2giXqtt/M5CoUilDTV1HNycvj73/+OyWRixIgR/PTTT81jnnvuOZYtW0ZJSUmH5zKbj7Qx63S65uc6nQ6v1wvAjTfeyPz589m0aRN//OMfmxf6vPDCCzz66KPk5+czZswYysrKAEhPT6e6uprt27cH9ecONhEp6ulJDvwS9pV1XDIw9utHXOUOQJKfW949wUUYDpNWfql1K1FX9DzOOeccXC4Xzz9/ZE1jXV1wSoXV1dX069cPj8fDW2+91fz6rl27mDBhAg8//DCJiYnk52sd2wMGDODDDz/k+uuvZ8uWLUGJIRREqKg3dcB0XDIQRiO2pBhidVUU5FV0R2gRh81oA6DWo0Rd0fMQQrB48WJWrFjBoEGDGD9+PDfccANPPvnkCZ/7kUceYcKECZx//vlkZmY2v/673/2OUaNGMXLkSCZPnszo0aObjw0bNoy33nqLK6+8kl27dp1wDKFAaCXx7ic7O1t2deejapeHUX/6gt9NG8bcqUM6HLvvxpvI5RT22Mdwy1/OwmRRHmYt8fl9jHljDLePvp1fjvlluMNRhInc3FyysrLCHYbiKNr6dxFCrJNSZrf3nojM1KMsRvpEmwPqVTc6U4g5lIPfJync2fnk6smGXqfHarCqPnWFopcQkaIOkJ7oCKit0eRMJWr/evQGQX6eqqu3hd1oV+UXRcSxdOlSxowZ0+rRnatMeyoRW4tIT7Lz7w0HkVJ2aAFgTHWi93vo099EQa6qq7eFw+hQoq6IOKZNm9bchqg4QsRm6oOTHFS5vJTVujscZ0rVHAz6xDRQdqCGuqqOx5+M2Iw2JeoKRS8hYkW9qQNmV3HHJRhjo6gniSIACrapEszRqExdoeg9RKyoD07S+qt3l3YsRvrYWHR2O47yPZhtBlWCaQOb0aYmShWKXkLEinr/WCumAIy9hBAYU1PxFuSTMiyO/LxywtXG2VNRmbpC0XuIWFHX6wTpifaA2hpNqU7cBfmkZsZRU95AZUkgO2ycPKjuF0VPQK/XM2bMGEaOHMmVV17ZvHK0qKiI2bNnk56ezrhx45g4cSKLFi0Kc7Sd43A4wnLdiBV10OrqgbQ1Gp2pePILSMmMA1CrS49CibqiJ6D81INDxLY0gtarvnRLEW6vH5Oh/c8nY6oT2dCAnRoc8WYKcssZOTmlGyPt2TiMDjx+D26fG5PeFO5wFOHms3vh0KbgnrPvKLjgiYCHn6if+pIlS6irq2PXrl1cfvnlPPXUUwAsXLiQxx57DCklM2bMaLYbcDgczJ07l//85z/ExcXx2GOPcc8997B//36eeeYZLrnkEvbu3ct1111Hba2WAM2fP59JkyZRWFjI1VdfTVVVFV6vl+eff56zzjqrOZ7S0lIuvvhi7r//fmbMmHFcv7auENGZ+uBkOz6/ZH95x1lmU1uj98ABUjPjKdhWgd+v6upNNPm/qMlSRU8gGH7qOTk5vPPOO2zatIl33nmH/Px8Dh48yO9//3uWL19OTk4Oa9asYfHixQDU1tYyZcoU1q1bR1RUFPfffz9ffvklixYt4sEHHwQgOTmZL7/8kp9++ol33nmHX/3qVwAsWLCAadOmkZOTw4YNG5rtg0ErHc2YMYOHH364WwQdekGmDrCzuJYhyVHtjjM6nQB48vNxZk0gd1UhpfnVJA/ouZ7I3UnTRhm1nlriLfFhjkYRdo4jow4mwfRTP/fcc4mJ0fbcHT58OPv27aOsrIwpU6aQlJQEwLXXXss333zDZZddhslkYvr06YDmuW42mzEajYwaNYq9e/cC4PF4mDdvHjk5Oej1+mYL3tNOO42bb74Zj8fDZZdd1vwzeDwezj33XJ577jnOPvvs4P2iOiGiM/Vmt8bSTnrVU1JACNz5BTiHaaKlrHiP0FLUFYpwESo/db1ej9fr7bDrzWg0Nq9Mb89//W9/+xt9+vRhw4YNrF27FrdbW8g4efJkvvnmG1JSUrjuuuuad1YyGAyMGzeOpUuXduG30XUiWtSjLEaSojo39tKZTBj69sWTn48t2kRCikNNlrZA2e8qeirB9FOfMGECK1asoLS0FJ/Px8KFC48rg66srKRfv37odDreeOMNfD5tk559+/aRnJzMrbfeyi233NL8ISSE4JVXXiEvL48nnui+u5+IFnXQdkEKZBNqk9OJu3HG3JkVR+HOSrzujndOOllQmbqipxJMP/V+/frx+OOPM3XqVEaPHs2pp57KpZdeGvD7b7/9dl577TVOP/10tm/fjt2uVQq+/vprxowZw9ixY/nggw/49a9/3fwevV7P22+/zVdffcU//vGP4465K0Skn3pL/rBoE59sLCTnwfM7NPY6+If7qF25koxvVrBvcxkfz9/AJb8eQ2qWqiHvPrybSz+6lCfPepIL0y8MdziKMKD81HsmJ42fekvSE+1U1nso79TYy4m3uBi/y0W/ITHo9IICZcULaH3qoPYpVSh6AxEv6oOTA/OAMTq1tkbPgQOYLAb6pseQr3xgALVPqSIyUX7qbRPRLY0AgxvbGneX1HDawPZLKaZUra3RnZ+PefBgnJlxrP54D64aDxaHsVti7alYDVZAZeqKyEL5qbdNxGfqKXGasdeuTjpgmix4PfnaZGlqVjxIKNimsnWd0GE32qlxq8VHCkWkE/GirtcJBibYOu2A0cfHI2w2PAX5ACQPiMJo0au6eiN2g506b9daxRQKRc8h4kUdNG/1znrVhRBaW2Njpq7T60gZGke+6lcHwG5SmbpC0RvoFaKenmRnX3kdbq+/w3HG1FQ8+fnNz1Oz4qgqqaeqVFnxKk91haJ30DtEPdHRaOzVcfmgaQFSU2++M1ObWFWrS9U+pYrw0x1+6pMmTQK0BUMXXXRR0GLvCqHyW+8dot7kAdNJXd2Ymoqsr8dXVgZAXF8b9hgT+aqujsPoUC6NirDSHX7qq1atCnLUPY+Ib2kESA9wv9KWbY2GxESEEDiz4tm3uQzplwhd+ytSezt2o506j5ooVcCTq58krzwvqOfMjM/k9+N/H/D4E/VTX7RoEQ0NDezZs4fZs2fzxz/+EdCy45oaLXmpqqri8ssvZ9u2bUyePJl//OMf6HRt57mR5LfeKzL1GKuRRIeZXcWdZ+oAnhaf8qmZcbhqPJQeOLmzVLvRrjJ1RY8gGH7qq1ev5q233iInJ4f33nuPtixJVq9ezV/+8hc2bdrErl27+PDDD9s9XyT5rQeUqQshpgPPAnrgZSnlE0cdTwNeA2Ibx9wrpfw0KBEGSHqSvfNVpSnabkfuFpOlzXX13AqSUtv3ZO/tNG1pJ6Xs0ENH0fs5now6mATTT/38888nISEBgJkzZ7Jy5Uqys1vbpYwfP5709HQAZs2axcqVK7niiivaPF8k+a13mqkLIfTAc8AFwHBglhBi+FHD7gfelVKOBa4BuseOrAVaW2PHmabObMbQp0/zAiQAe6yZuH72k75f3W6045M+GnwN4Q5FcZISTD/1oxOTthKVQMY0EUl+64Fk6uOBnVLK3QBCiLeBS4GtLcZIoGkboRjgYDCDDITBSXYq6jRjr3h7+/tsGlOdrdoaQSvBbPyqgOfnfdXpdYQQTL56KMPP7H/CMfckmux3azw1WAyWMEejUGicc845/OEPf+D555/nl7/8JRCYn/qXX35JeXk5VquVxYsX88orrxwzZvXq1ezZs4cBAwbwzjvvMGfOnBOKtbKyEqfTiU6n47XXXmvlt56SksKtt95KbW0tP/30E9dff32z3/qVV17JE088wb333ntC128iEFFPAVqqYAEw4agxfwK+EELcAdiB89o6kRBiDjAHIC0t7Xhj7ZCWHTDx9g48YJyp1P7wQ6vXxpyfhtGiR3bc5g7AjjVFbF99qNeJepNTY52nDqxhDkahaKTJT/2uu+7iqaeeIikpCbvd3qmf+plnnsl1113Hzp07mT179jGlF4CJEydy7733smnTJiZPnnzCZmC33347P//5z3nvvfeYOnVqK7/1p59+GqPRiMPhaM7U4Yjf+sUXX0x0dDS33377CcUAgYl6W/ckR5uwzwJelVL+RQgxEXhDCDFSytYyKaV8CXgJND/1rgTcHunNxl61ZHdg7GVMdeL9qAh/QwO6xluoqHgLp186OKDrSL9kw1f5eBp8GM36Ew+8h9Ak6mqyVBEumrpSjqZfv368/fbbx3Wu5ORk5s+f3+41pkyZwpQpU7oU25/+9Kc2j2VkZLBx48bm1x9//HEAbrjhBm644YZ2z2kymYJaggmk+6UASG3x3Mmx5ZVbgHcBpJTfAxYgMRgBBoozzopJr2NXJ/uVmlJTQUo8B7pWIXJmxeH3Sgp3Hu7S+3sqzZ7qagGSQhHRBJKprwEyhBCDgANoE6GzjxqzHzgXeFUIkYUm6h3PZAQZg17HgAQbu4oD9FUvyMecPui4r9NvSCw6gyA/r4K0EQldirUnora0U0QaS5cu5fe/b92pM2jQIBYtWsSNN97YpXNOmDCBhobWzQJvvPEGo0aN6mqY3U6noi6l9Aoh5gFL0doVX5FSbhFCPAyslVIuAe4G/imEuAutNHOjDMM+eelJdnZ00qvecgFSVzCa9PQbHNPrumVU+UURaYTCT/3HH38M6vnCQUB96o09558e9dqDLb7fCpwR3NCOn8FJDpblFuPx+THq264s6RMTERZLq7bG48U5LJ4fl+ymvsaN1dF+p00k0WqiVKFQRCy9YkVpE+lJDrydGHsJITA6U3AXdC1TB62uDr3LCExl6gpF76CXiXpTW2MnHjDO1BPK1JPTojBZDb1K1K0GKzqhU57qCkWE06tEveV+pR3R5Kve1bK/tsFGbK+qqwsh1O5HirDSZL07evRoTj311FaOitu3b+fCCy9kyJAhZGVlcdVVV1FUVATAypUrGT9+PJmZmQwbNoznnnsuXD9Cj6BXuDQ2EWMzkugwdZ6ppzrx19Xhq6jAEN9+T3tHpGbFs2dDKZUl9cQk9Y7VOmr3I0U4abIJAK2z5b//+79ZsWIFLpeLGTNm8Ne//pWLL74YgK+++oqSkhKklMyePZvFixdz6qmnUlpayrRp0+jfv/8JLyaKVHpVpg7aIqRdnWXqTW2NXeyAAXBmNtXVe0+2rjJ1RU+hqqqKuDjtb2zBggVMnDixWdABpk6dysiRI3nuuee48cYbm50cExMTeeqpp3j66afDEndPoFdl6qDV1b/YWtThmCNtjQVYR4/u0nVi+9hwxJnJz61gxFkpXTpHT0Nl6gqAQ489RkNucP3UzVmZ9P3DHzoc0+TS6HK5KCwsZPny5QBs3ryZcePGtfmeLVu2HLNaMzs7m61bt7Y5/mSg12Xqg5MclNe6qah1tzvG6NRE3XMCHTBCCJyZcRRsK0f6u70lPyTYDXZqvWrxkSI8NJVf8vLy+Pzzz7n++us7nfdSVtHH0iszdYDdpTWMa8fYS2e1ok9K7PICpCacmfHkfX+I0oIaktIi34vdYXJQXFcc7jAUYaazjLo7mDhxIqWlpZSUlDBixAhWrFjR5rgRI0awdu1aLrnkkubX1q1b16aB18lCr8vUm7a22xXitkY4UlfPz+0ddXW1+5Gip5CXl4fP5yMhIYHZs2ezatUqPvnkk+bjn3/+OZs2bWLu3Lm8+uqrzROsZWVl3HfffTzwwAPhCj3s9LpMPTXOilEvOu2AMaY6qWtji6vjwR5jJr6/tsHGqdMGnNC5egJqn1JFOGm585GUktdeew29Xo/VauXjjz/mzjvv5M4778RoNHLKKafw7LPP0qdPH958803mzJlDZWUle/fu5dVXXw3qTkKRRq8Tdc3Yy95pr7rJmUrVvz9Gut0IU9eX+qdmxrP52wN4PT4Mxsi24rUbtZq6qlMqwkHTphJtkZmZyeeff97mscmTJ7N69WpA2xnpscceY/r06c3dMycbva78ApCeaO+8rbHJgvfgiW3S5MyKw+fxc2hX5QmdpydgN9rxSz/13vpwh6JQdIm5c+eyadOmk1bQobeKepKD/eV1eH3tb2XUsq3xROifEYtOp1nxRjrKflehiHx6pagPTrLj8UnyK9rPOI2pR3zVTwSTxUCf9GgKesFkqc1oA5Spl0IRyfRKUW/qgOmorm5ISkKYTCecqYPW2li8vxpXreeEzxVOmjJ1NVmqUEQuvVLUBzf2qndUVxc6HUan84SsAppIzYwDCQe2R3YJRtnvKhSRT68U9VibiXh758ZexlQn7oITz9STB0VjNOspyO0doq5q6gpF5NIrRR20bD0wX/WuW/A2oW+04s2PcHMvNVGqCCdN1rsjR47kyiuvpK5OKwMWFBRw6aWXkpGRQXp6OvPmzWveR/Trr7/moosuCmfYPY5eK+oBuTWmOvHX1OA7fPiEr+fMjKeyuJ6qsshtB1QTpYpw0uT9snnzZkwmEy+88AJSSmbOnMlll13Gjh072LFjB/X19dxzzz3hDrfH0ntFPclOWa2byrr2Jy9NzR0wQZgs7QVb3KlMXdFTOOuss9i5cyfLly/HYrFw0003AVo2/7e//Y3XX3+dmhqVfLRFr1tR2sTgJg+Y0hpOTWt7IUJLX3XrqFEndL34fnZs0SYK8ioYfkb/EzpXuDDrzRiEQYn6Sc63726nND+4gpmY6uCsq4YGNNbr9fLZZ58xffp0tmzZcoztbnR0NAMHDmTnzp1BjbG30Kszdeh4v1KTU/NBD0ZboxACZ1YcBXmRa8UrhMBmtClRV4SFJu+X7Oxs0tLSuOWWW9q1rDjRebDeTK/N1FPjbZgNOt74YR9nZSTSJ9pyzBid3Y4+IeGEFyA1XzMznu0/FlF2sJZEpyMo5+xuHEaHEvWTnEAz6mDTcju7JkaMGMEHH3zQ6rWqqiqKiooYNmwYP/74Y3eGGBH02kzdqNfx9JWj2X6omguf/ZZvd5S0Oc7kdAYlU4fescWdzWhTux8pegznnnsudXV1vP7664Bm+nX33Xczb948rNbesTdwsOm1og5wyej+LJl3BvF2E9e/spq/frEN31GlEWNqalAWIAE44izE9bWRH8H96g6jQ+1+pOgxCCFYtGgR77//PhkZGSQkJKDT6bjvvvuaxyxbtgyn09n8+P7778MYcfjpteWXJjL6RPHRvDN4YPEW/nf5TtbsreDZWWNIjtLKMcZUJ1WffYb0eBBG4wlfz5kZT+6qg/i8fvSGyPvMtJvsVLoi33FSEXm0182SmprKkiVLAFi1ahWzZs1i3bp1jBs3jilTplBfH7ltxKEg8lSnC9hMBv5y1WievuIU1udXcOGzK1m1sxTQFiDh8+E5dCgo13JmxuF1+ynaE5nCqPYpVfRkJk2axL59+9rdiFpxkoh6E1dmp/LR3DOJsRq49l8/8sx/tqNv2oQ6SCWYlGFxCEHEWvE6TA5q3UrUFYpI5aQSdYBhfaNYMu9MLhuTwjP/2cE932kZe7AmS81WA8kDoyPWB8ZmsKlM/SRFtQn2LLr673HSiTqA3Wzgr1eN5omZo/i6TOLRGdi7eXvQzp+aFU/R3irc9d6gnbO7cJi0lka/bH+DEUXvw2KxUFZWpoS9hyClpKysDIvl2FbszghoolQIMR14FtADL0spn2hjzFXAnwAJbJBSzj7uaLoRIQTXjE9jdGosRf+JZ8eqzXj3lnPawPgTPrdzWBxrP93LgR2HGXRKYhCi7T5aeqo7TJHZa684fpxOJwUFBZSUtN36q+h+LBYLzsby8PHQqagLIfTAc8D5QAGwRgixREq5tcWYDOC/gTOklBVCiOTjjiRMZPWLxjwiA9f2fH755k/8+44z6BdzYv2vfdNjMBh1FOSWR5yoN5l61XpqlaifRBiNRgYNGhTuMBRBIJBMfTywU0q5G0AI8TZwKbC1xZhbgeeklBUAUsriYAcaSmwD0xiQk8M5m5fxxt0ruWHSQIz6Y5cmAyB0RE2bhrFP+59beqOO/hmxETlZqky9FIrIJhBRTwFatoYUABOOGjMUQAjxHVqJ5k9Sys+PPpEQYg4wByAtLa0r8YYEyymnwIKF3JSzGIDylR2Pd+/fT9/77+twTP+hsfyweDcN9V7M1shZDqA2ylAoIptA1KatlPXo2RQDkAFMAZzAt0KIkVLKVkblUsqXgJcAsrOze8yMTOxllxF13vng9/GPr3byword3HvBMGZPGHDM2Pw5/4UrN7fzcyZrZYyq0nqSUqOCHnOoUFvaKRSRTSDdLwVAaovnTuBgG2M+klJ6pJR7gG1oIh8x6B129NHRzL14LBNHD+RPX+WzusSDPjq61cMyYgQNublIf8fdIdGJWl2+qjSyVrupTF2hiGwCEfU1QIYQYpAQwgRcAyw5asxiYCqAECIRrRyzO5iBdhc6neBvV49mYIKNuQt+oqCirtVxS1Ym/rq6ThcrRSc1inqJK2SxhgIl6gpFZNOpqEspvcA8YCmQC7wrpdwihHhYCHFJ47ClQJkQYivwFfA7KWVZqIIONVEWI/+8PhuP189/vbGOerev+Zg5MwsAV25eh+cwWw2Y7YaIy9SbJkpV+UWhCA1Vn36K3xW6ZC+gxUdSyk+llEOllIOllH9ufO1BKeWSxu+llPI3UsrhUspRUsq3QxZxN5Ge5ODZWWPYWljFvR9ubF6UYc4YAno9rrzO6+oxidaIE/WmTL3OU9fJSIVCcbxUfb6UA7+5m/I33gjZNU7KFaWBck5mH377s2F8lHOQl7/dA4DObMacnk5DJ5l6fnkd0m6gMsJE3aQ3YdQZVaauUAQZ9759FN53H5bRp5Bwww0hu07k9NqFidunDGbLwUoe/yyXzH5RnJWRhGV4FrXf/9BqnM8vyck/zLLcIpblFrOtqJqzXUZO9xjx+yU6XTt97z0Qu9GuauoKRRDxu1wU3HkXwmDA+be/IUymkF1LiXonCCF4+orR7CquZd6C9fx73pnYM7Oo/GgJVQeLWFUuWZZbxPK8Yspq3eh1gtMGxvFfk9P5/st9+H2S2sMNRMUfv4dDuFCirlAEl6LHHqchNxfnC89j7B/ajemVqAeA3WzgpevHccn875jzxlpui4plGHDLgwtZk5hBtMXAlGHJnJuVzJShycTYjEgpueyHA1APVSX1ESXqDqNDlV8UiiBRuWQJh999l4RbbyVqypSQX0+JeoAMSLDz91ljufH/VvPAfjfvAlfE1PGbW08ne2AcRn3r6QkhBDmAtLwAACAASURBVKOGJkBxBeXFdaQMiwtP4F3AbrSriVKFIgg07NxJ4R//hDV7HEm//lW3XFNNlB4Hk4cmsej2M1h074UY+vXlTFHBxMEJxwh6E2eN7osfybZdkeUBYzfaVaauUJwg/ro6Cn59JzqbjZS//BVh6J4cWmXqx8no1FgA8rOGd2oXcObQJNbqwLu/qjtCCxp2o5391fvDHYZCEbFIKTn00EO4d+8m7ZV/dWgAGGxUpt5FLJmZuPfswd/Bprd2swHhMFBdHnmrStVEqULRdQ6//z6VHy0hcd5c7BMnduu1lah3EXNWJvj9NOzY0eG4hD42LA2SvaWRI5IOo0OJukLRRVy5uRQ98ij2SZNIvO22br9+7xT1mmJY/ig8NRj+ng1fPAD7VoEveNvLWbICswvISI/FLgXLNh0K2rVDjd1op95bj8/v63ywQqFoxlddTcGdd6KPjaX/008h9Ppuj6F31dSL8+D7+bDxHfB5YOh08NbDD8/Dqv8Faxxk/Ex7fci5YInp8qWMKSnooqI6tQtIS4shl3x+3FTMLVMHd/l63UmzqZe3lmhTdJijUSgiAyklhfc/gKfgAANeexVDQkJY4oh8UZcS9n4Lq+bDjqVgsMDY62DiXEhoFFFXFexaBts+18ZsfAd0Rhh4Bgy9AIZNh7iBx3VZIQSWzEwatnYs6tGJWn/6/vxKahq8OMw9/1feLOpuJeoKRaBUvPkW1UuXkvzbu7FlZ4ctjp6vMO3h88DWj7QMvHAD2BJhyh/gtFvAftS+oJZoGHG59vB5oWA1bPsMtn8On/9ee6RkwzULIKpPwCGYszI5/N77SJ+v3dusJl91h0ewckcJ00f26/KP3F3YTcp+V6E4Huo3bqToqadwTJlC/M03hzWWyBN1VxX89Dr8+AJU5kNCBlz8LJxyNRgD2DBab4ABk7THzx6Bsl2w7VP46nF443K48WOwxQcUiiUzC1lfj3vffszpbW/aa7EbMVkNJAkfy/OKI0LUlf2uQhE4nsJCCubOw5iURP8nHkfowjtVGXkTpav+Dl/cB7EDYNY7MHc1jLsxMEFvi4TBMOkOmLUAynbAmz/XPjgCwJKVCUBDJ3X16EQLAy1mlueV4Pf3mF382kXZ7yoUgeGrqSX/tl/ir6/H+cLz6GNjwx1SBIr6+Fvh1uVw0ydaLTxYn4rpU+DK17RSzsJrwN25oJkHDwajsdMOmJhEK7F+QWlNA5sPVgYn3hCi9ilVKDpHer0c+M1dNOzcScozz2AZOjTcIQGRKOqOZEgZF5pzZ14IM1/S2h/fvR687g6HC5MJ85Ahna4sjU604q/2ogOW5RYHMeDQoLa0Uyg6RkpJ0WOPUfvNt/R98EEcZ54R7pCaiTxRDzWjrtBq9Du/hA9u6bS33ZKZiSuv40w9OsmK3yc5vV8sX23r+aLeVFNXoq5QtE3F669TsWAh8TffTNzVV4U7nFYoUW+LcTfAtMcgdwksmQd+f7tDLVmZ+EpL8ZaUtDumqa3xjL4xbCyopLiq59gGfLzxINe89H2rWr/NaANU+UWhaIvq5cspeuJJos4/j+Tf3h3ucI5BiXp7TJyrtUhuWAif3aP1w7eBOVObLO0oW29qa8yK1sTy623tfwB0J1JK5i/fyQ+7yymtaWh+3agzYtab1USpQnEU9Vu2cODu32IZOZL+Tz0V9k6Xtuh5EfUkzr5H64xZ809Y9lCbQyxNot7BZGlUvAUhwO6B/jEWluUVhSTc42V9/mHyDlUDkF/R2phM2e8qFK3xFBZScNsv0cfFkvqP59BZu9hxF2KUqHeEEHD+I5B9M6z8G3zzP8cM0UdHY3Q6ceVubfc0eoMOR5yFqrJ6pmYms3JHKQ3e8PuqvL16P6Jx69SCitZZuXJqVCiO0Ny6WFdH6gsvYEhKCndI7aJEvTOEgAv/AqOuguWPwI8vHjPEkpVJQydtjdFJFqpKXJyblUyt28fqPeWhijggql0e/r2hkEtGa/slFhyVqSunRoVCQ3q9HLj7Nz2udbE9lKgHgk4Hlz0PmRdp9fWtH7U6bM7MxL1vH/7a9kUwOtFKVWk9E9MTMRt0LM8LbxfMRzkHqff4uOmMQSQ6TMdk6jajTZVfFAqg6PEnqF3xDX0feADHWWeGO5xOUaIeKHoDXPEK9BkJyx5p1RFjycoCKXFt397u26MTrdRVuTEAZwxJZHleMbKdydfuYOHq/WT1i2a0M4aUOFubmbqaKFVEAlJKvBUVuPLycG3bjnv/fjxFxfiqqpBud8B/Z9LrxVdTi7e0FHfBARp27qT0xZeoeOst4m+6ibhrrg7xTxIcIs/7JZwYzHDmXVr/+vbPtcVKHJksbcjLwzZ2bJtvjWnsgKkq1erqy/OK2V1ay+AkR/fE3oJNBZVsOVjFw5eOQAhBapyVzQdar3RVE6WKnoKvqgpP4SG8hwrxFB7Cc6gQb+EhPIcavz9UhGxoaP8EOh06iwVhtaIzmxFWK0Knw+9yIV0u/I0PPJ423x51/nkk/+63Ifrpgo8S9eNl+GXwn4fgu2ebRd3Qrx/6mBhcHdjwRjeLuotzMpN5AFieWxwWUV+4Zj8Wo45Lx6QA4IyzsXTLIfx+iU6nzZyqiVJFuJE+H4UPPEjlhx+2PqDXY0hOxti3L9YRIzCccy7Gfn0xJPcBIZCuevz1LmSDC3+9C7+rHulq0L7Wu/A3NIDPi7BYG8Xegs7c+NViRVjM6CxWdFYLOocD+4QJPbJ1sT2UqB8veoPWw/757yF/NaSORwiBOSur4171JG0BUlVJPaNPSSSzbxTL84q5dXJ6d0UOQG2Dl4/WH2DGqP7EWI0AOOOseHySomoX/WIarYLVRKkijEi/n8L77qdy8WLirr0WW/Y4DH37YuzXD0NiIsKgpKs9Iufjpycx9hdgidWy9UYsmZk0bN+O9LZtK2CxGzFa9FSVarXrczKTWbO3nCpX27d8oeLjjQepdfuYPSG1+bXUeG1RVMu6us1oo8HXgMffvfEpFNLvp/DBB6lcvJjEX91B3wfuJ/qCC7CNHYuxb18l6J2gRL0rmB1w2v+DvE+gdCegtTXKhgbce/e2+RYhRHMHDGii7vVLvt1e2l1RA7BgdT4ZyQ5OTYtrfs0Zp2XnLTtgmvxf1GSpojuRfj+H/vQQle9/QOLtvyTp9tvDHVLEoUS9q0z4L9CbtD1RAXNm5xtRxyRaqSzVfF/GpsURazN26+rSrQer2JB/mGvGpyGaVh0BKbGaqOeXH8nUlf2uoruRUlL06KMcfvddEubMIfGOO8IdUkQSkKgLIaYLIbYJIXYKIe7tYNwVQggphAjfBn3dhSMZRl8DOQugphhz+iCEydShDW90ooWq0nqklOh1gilDk/h6Wwm+bto44+01+zEZdMwcm9LqdYtRT3KUuVWmrux3Fd2JlJKixx/XnA9vuZmku+5slXgoAqdTURdC6IHngAuA4cAsIcTwNsZFAb8Cfgx2kD2WSXeAzw2rX0IYjZgzMjrcBSk60YrP46euSvNpPyerD+W1bjYUHA55qPVuH4vWH+CCkX2Js5uOOe6Ms7aqqSv7XUV3IaWk+MmnqHj9DeJvuJ7k3/5WCfoJEMiMw3hgp5RyN4AQ4m3gUuBos5NHgKeAyGnoPFESMyBzBqx5Gc68C3NWJjXLliOlbPM/ZXRSY1tjST32GDNnZySh1wmW5xa3qnGHgk83FVLt8jJrfFqbx1Pjbfy0v6L5ebP9rluVXxShQ0pJyV/+QvmrrxJ37bUk33tvq7+dqtJ6tn53EE+DD6/Hj9ftw+f243Fr3ze91vQ9PX+3SAAmzhxM5umh2a84EFFPAfJbPC8AJrQcIIQYC6RKKT8WQrQr6kKIOcAcgLS0tsUl4pj0K8j7GNa/iSUzi8r3P8BbXIyxT59jhrZcgNRvSCwxNiPjBsSxPK+Y304bFtIwF67eT3qinQmD2t5U2xln5ZONhXh9fgx63ZFM3asydUVokFJS8uyzlL38L2JnXUOf++9rJehSSpb+czPF+6sxmfXoTXqMJh0Gkx6DUftqsRswxJoxmHTojTqELjIy/Kh4S8jOHYiot/Vbav48FELogL8BN3Z2IinlS8BLANnZ2RHymdoJaRMgdQJ8Px/LpH8B4MrNbVPUo+ItIGieLAWtC+aJz/IorKxv7hEPNjuKqlm7r4I/XJjZ7m2tM86G1y8pqm4gJdaKw9Qo6m4l6orQUPrcPyh74UVir7yCvg88cMz/zT0bSineV80512eSNal/mKKMPAKZKC0AUls8dwIHWzyPAkYCXwsh9gKnA0tOisnSJib9Cg7vx8wOABramSzVG3U4Ys3NbY0A52YmA/Diit24PKGx4124Oh+jXvDzU53tjkmN08ot+eXaZGlT+UXV1BWhoPSFFymdP5+Yyy+n70MPHbNiU/olqz/eQ0ySlWET+oYpysgkEFFfA2QIIQYJIUzANcCSpoNSykopZaKUcqCUciDwA3CJlHJtSCLuiQy7EBKGoF//Isa0tA7bGlv2qgMMSXZw8ej+vLpqL1Oe/poFP+7H42t/+7zjxeXx8eH6An42oi8JDnO74470qmux2Q2q+0URGjxFRZQ88wzRF15Iv0cfaXMJ/q71JZQV1HDaRYPQ6VXn9fHQ6W9LSukF5gFLgVzgXSnlFiHEw0KIS0IdYESg02mdMIUbsKQldmIXYKWq5IioCyH4+6yxvDPndFLirPxh0SbO/+sKlmw42Grf0K6ydMshDtd5mHVax3MY/WK13ZmaMnW9To/VYFWirgg69evXAxB/040Ivf6Y4/7GLD2ur42M044tYyo6JqCPQCnlp1LKoVLKwVLKPze+9qCUckkbY6ecVFl6E6dcA/ZkLIZ8PPv346tpu2skJtFCbaUbr7t1qWVCegLv3zaRf92QjcWo51cL1zPj7ytZnld0Qha9C1fvJzXeyqTBCR2OMxv09I22tGprVE6NilBQv349wmJpdjc9mp3riqgorNWy9AiZ+OxJqPuaYGG0wIQ5mP1alt6wbVubw5rdGstcxxwTQnBuVh8+/dVZPHvNGGobvNz86lquevH7Lu2UtLukhh92l3PNaWkB/XFoveqtrQJUpq4INnU5OVhGjkAYjccc8/v8rPl4LwkpdoacmhyG6CIfJerBJPsWLEnaf9T2bHijW7Q1todOJ7h0TArL7j6bRy8byb6yOq568Xtu/L/V5OQfDjhzf2dNPnqd4Mpx7U+QtsR51GYZNqNNiboiqPhdLlxbc9vdd2D76iIOF9Ux/qL0iGlP7GkoUQ8mtngMZ1yL3uzHtXFdm0MCEfUmjHodvzh9ACt+N5V7L8hk/f7DXPbcd5z252X8+u31vLsmnwOH2z6P2+vn/XUFnJeVTHJ0YD2xqXFWCivrmydqVaauCDauLVvA48E6Zswxx3w+P2s+2UNiqoNBYxLDEF3vQHlYBhkxcS6WuPdpWP9Dm8etUUYMZj1VJceWX9rDatJz29mDmTU+jaWbD/HdrlK+21nKRzlaZ+mgRDuTBidw5pBEJg5OINZm4sutRZTVurmmnRWkbeGMs+GXcKjSRWq8DZvRRmFNYcDvVyg6oz4nB6BNUd/2/SGqSl3MuP0UZRNwAihRDzZxAzAPGUTFqgPI6jJEVOsJSiEEMYkWKgPI1I8mxmrkqtNSueq0VKSUbC+qYeXOUlbtLGXx+gO89eN+hICR/WOoafCSEmtlckZSwOdvamvML68jNd6Gw+hQE6WKoFK3fj3GtDQMCa3/LnweP2s+3UOfQdEMGNXxpL6iY5SohwDL2TORK+fT8MkzWK555Jjj0YlWKkuOX9RbIoRgWN8ohvWN4pYzB+Hx+dmQf5jvdpbx3c5S8g5Vcc+0TPTHUZc8erMMtaWdIphIKanP2YDjjEnHHMtddZCa8gam/qL9Vc+KwFCiHgIsE6cB83F99S6WGXMhqvWKuOhEK/m55e0af3UFo15H9sB4sgfG8+vzMvD55XEJOkDfGAs6AfmNHTBK1BXBxFNQgK+0FOtRk6Rej4+1n+2j3+AYUrPa9iZSBI6aKA0BpkGDEGYTDcUeeO0SqCludTw60YrX7ae+OnRbxR2voIP2wdAv5ogFr8PowOP34Pa5gx2e4iSkadHR0fX0Ld8cpPZwA+MvSVdZehBQoh4ChF6PeVgmLtNoqMzXhL32yLZ10YmNm1B3oa4ealr2qiv/F0Uwqc/JQWezYc7IaH7N4/axbuk+UobF4hwWWvvpkwUl6iHCkpmJa28hctbbULEXXr8U6rQFRDGNvuonWlcPBc44W/O2dk32u2qyVBEM6tbnYBl9SitrgM1fH6C+ys34i9PDGFnvQol6iDBnDsNfWYnXOhRmLYTSHfD6JVBXTlRCz83UU+OtFFW7aPD61JZ2iqDhr62lYdu2VouO3C4vP32xj9Th8fQfEhvG6HoXStRDhGXoUAAaduyAwVPhmgVQsg3euByDtxr7URa8PQVnnA0pofCwS4m6ImjUb9oEfn+rSdJNXxfgqvEwQWXpQUWJeohoqhs2bN+uvZBxHlz9JhRtgTdnEh1vpKo08AVI3UVzr3pFndqnVBE0midJTzkFgIZ6L+u/2M/AUQn0GRQdztB6HUrUQ4Q+JgZDnz64mkQdYOg0uOp1KNxATOUqqkp6nli27FVXmboiWNTl5GAaMhh9TAwAG5bl01DnVbX0EKBEPYSYhw6lYfuO1i9mXghX/B/RDVupOezGV1sVnuDaoU+UGYNOUFBR1yzqaqJUcSJIv5/6nA3NrYwNdR42/Gc/6WOSSEqLCnN0vQ8l6iHEPDQD965dSK+39YHhlxA98RJAUP3a7dCD9gE16HX0i7WQX94iU+9B8SkiD/fevfgrK5snSTd/cwC3y0f2hQPDG1gvRYl6CDFnZCA9Htz79h1zLHrMWQBU5h+AhdeAu+6YMeEiNc5GQUXdkT51rxJ1RddprqePHYvX42PD8gLShserLD1EKFEPIc0dMC3r6o00L0AaeSfs+RaWzIMT2OEomDjjrORX1KMTOm33I7cqvyi6Tt369ehiYjANHEje94eor3IzdtqAcIfVa1GiHkJMgweDTqe1NR6FLdqEwaijyjwczrkfNn8AP70ehiiPJTXORkl1Ay6PD7vBTp2359xFKCKP+pwcrGNGIxHkfLmf5AFRpAxVfemhQol6CNGZzZgGDGjdAdOIEIKoRKvW1njmbyB9Cnz2eyhue8ek7sQZr7U1Hjhcj92kMnVF1/FVVuLeuQvbmDHsXl9CZUk9p04bcHJ7vFQXgd8fstMrUQ8xbXbANNLsq67TweUvgdkB790Y9vq6M06rpeeX12E32FVNPYJw5+cfOzEfRuo3bgTAMmYMPy3dR0yylUFjAvf471X4PLDq7/D3U2F96O7KlaiHGPPQDDz5+fjrjhXq6EQrVaX12p6jUX3g8hehJA8+vzcMkR4hNa5Fr7rJrrpfejjS76d62TL2/uIX7Dr/Z5S9/HK4Q2qmfv160OkotwygZH81p/5sQECboPc69nwLL5wJX9wPA8+EQWeH7FJK1EOMeehQkJKGXbuOORadaMXj8uGqbbTgHXIunHkX/PSaVmMPE8lRZox6QUFFvdr9qAfjd7moePsddl84g4K58/AeLMQ0aBCH330PGcLb++Ohbv16zMOGsX5FEbYYE8Mm9O38TT0FKbUV4A3VXT9H9SH44FZ47SLw1MGst2H2OxA/KHhxHoXaJCPEWFrYBVhHjWp1LLrRrbGqxIXVYdJenHof7P0Olvwa+o+F+O5fcafTCVJireRX1BGbZqfOoyZKexLesjIqFiykYsECfBUVWEaOJOVvfyXq/POp/uILDvzmbup++AH7pGN3GOpOpM+Ha8NG/BfOpiCvgomXD0ZvjIA88vB+yFkIGxZoDqsGC2T8DEZcrq0KN9k7P4fPC6tfgq8eA18DTL4HzvoNGK0hD1+JeogxpqYiLJaO2xpL64/4X+iNcMW/tFu192+Gm78Ag6k7QwY0u4CCinpSBttVpt5DaNi9h/JXX6Xyo4+QDQ04pk4l4eabsGZnN088Os49F11MDIff/yDsot6wYwf+ujp2GkZisugZMTklrPF0iLsOcv8NOW/Cnm+01wadDWfcqTUvbF0MuUvAaIOh02HkTBhyXtsive97+ORuKN6ijbngKUgY3G0/ihL1ECP0esxDhrTZ1hid2OirfrRbY2waXDIf3r0Olj0E0/7cHaG2whln5cutRdiNWqYezK33FMeH5+BBDj36Z2qWL0eYTMRcdhnxN96AOf3Yuzid2UzMxRdz+N138R0+jD42fK2D9Tk51FmTyC8yMvZnKZitIZIbvw889eB1gckBRktg75MS8n+EnLdg8yJwV0PcQO1uefQ12t9hE9Mfh32rYMuHsHWJ9tXkgGEXagI/+BxwVcKXD8KGhRCTCle/BZkzoJv/bpSodwPmjAxqvv32mNeNJj22aFPbFrzDL4HTboXv58OgydptXzfijLNRWuPGpLPilV4afA1YDAH+sSiCSsn856hdtYrE228n7trZGBISOhwf+/OZVLz5JpUff0L8L67tpiiPpX79evIzZqDT6zjlnNTje7PfBwfWwbbPoGizJtrNj7pGEW98fvR2i0YbWOPB1viwxoMtocX38Y0llgVQvguMdq20MmY2DJjUtgjr9DDoLO1xwdOw91tN2HP/DZveBXMMILV4zrpbewRSpgkBStS7AfPQoVQuWoS3vBxDfOuNdZs6YNrkZ4/C/h9g0W1w20qI6b7b1yYLXo9bK/3UeGqUqIeJunVrsZ95Bkm/uiOg8ZasLCzDh3P4ww/CKuoVG3dQOOCXZE7siz3G3PkbGmpg13LY/jlsXwp1paAzQHIWmKPBGqtt4m60adm40aaVPwzWxq8WLduuK9ce9Y1fD+dr39cfBlqs2h5wJkz+LWRdorUTB4reoO2RMHgqzPgr7P4aNn+o1c6n/AEShxzvryqoKFHvBsxDj0yWGk4/vdWx6CQLhTsq236j0QJX/h+8eDZ8eCvc8G8tY+gGmnrV6xu0/yJ1njoI/RyP4ii8JSV49u0n7ppZx/W+mJ/PpOiRR3Ft3Ypl+PDju6jPC/UVUFd2RAyN1tZZr8nWcdxlZewhAz86xp6X1v7Aw/maiG/7TMt+fW6wxGgTk0OnazVpa5BKSH6f9rPUl2sfCMFIkvRGyDhfe/QQlKh3A0c8YHZgP1rUE63sWF2Ez+tHb2ijMyAxA2b8BRbfBiuegqn/3R0hk9q4qrTGpf0XUZOl4aFu3ToAbONO7Xig3w+F67VMt6GGGIebYoOOw8/8nr5XtvNen7sxqy07ktXWlUNDO0lGSwyWNkoc8WCOAgSV64s4kHIRA2MOELvxyWPf722AvSuhaJP2PH4wjJ+jCXna6ZpYBhudHuwJ2qMXE5CoCyGmA88CeuBlKeUTRx3/DfD/AC9QAtwspTzWmvAkRZ+YiD42loYdx3bAxCRakRKqy13EJreT/YyZBXtWwDdPNS5cOCvEEUOSw4zZoKOyVrszUBtlBEBDjdbXXHMIrHGta7mGAMoPbVC37ieE1YolK+vYg+467f/Fts+0ckXNIe11vRk9EJXioPKH7ST3/7HtGzy9sbUgx6cf+b6lYFtitUnIurIWZY0yqKs48n3RFu37Bu3Df+u2mXgdNsbpH4Ufdh97baHTWnbPfwSGXaAlL4qg0KmoCyH0wHPA+UABsEYIsURKubXFsPVAtpSyTgjxS+Ap4OpQBByJCCEwDx3apgdMUwdMVWl9+6IOcOH/QMEarc3xukXQd2SowgW0mFPirJRXa2KuRP0oqg/BoU1waGPj101QtotWNduWGO3HZrW2BK1GHO3USgHRKRDdv9UHQN26tVjHjEYYjUeu21Su2P11Y8dHlLZwbdgFMOT85kw09vvvqbrpZqpP+z9iZswI7e+jBT6Pn21zPyXBU0Dfx37otusqNALJ1McDO6WUuwGEEG8DlwLNoi6l/KrF+B+AXwQzyN6AeehQKj/8EOn3I3RHyixNor5zbTF1Ve723q4x7J/w3bPwv4/AhFuhz4hQhswYn5GqfSaGWk7j0Pp68g4UhvR6PRK/H2pLEJX7GWjNwVyeowl4bfGRMbEDoO8oOOVq7Wt0CrgOt5PVNj4/vE/76mqj1GFPhpgUfKa+NOTmknjxOPj6CU3MD65vvGYanHoDDJuuTfi1sZbBNmECxpQUKj/4oFtFPe+7Ahp0Nsb3VTfr4SAQUU8B8ls8LwAmdDD+FuCztg4IIeYAcwDS0jqYPOmFmDMy8NfV4Tl4EJPT2fy6PcaENcpI7qpCclcFIpo3aF8+Awito6O2XMIB/IKynbAsxNfr2TgYba/kzIxibRKv7yjt0WfEiU3kuWuh6iBUFkDVAag8AFUFUHmA+k07QIKtail8/TE4s+GcB7Te6OSsTvufhU5HzMzLKZ3/HO6CA5icoe+e8vslP32+B0f1fgZc2n0LbhRHCETU2/qf0+Y9phDiF0A20KZbjZTyJeAlgOzs7J6xI0Q30bIDpqWoC53gFw9PpL7GE/jJGmrgs99piyGyb4FJ87QaZZB568d9vPhNHvbBz3DzyJu5cuiVQb9G2Kgvh5JtmoFa8TYo3aYtCZeNnimmKEgaBsmZkDSMb9f0Z0fRZUya8z/BNaQy2bV6chs15bpnngH9y1gf3wRmk1ayOU5iL9dEvXLRIpLumBeMiDtkz4YSqg77GLH/S2xj/xLy6ymOJRBRLwBarhxwAgePHiSEOA+4DzhbStkQnPB6D+aMI7sgRZ1zTqtjJqsB03GttrPCzf+ET38L6x4DTx5c9nzgK+kCJC0tmkqdAWmpoN5WSUxSBPY0+v1QsedI3bupBl7d4q4oJhX6jYKxZx/JwGPTWmXCmfZi9v1zMwe2VZCadfzi2hXq167DMnw4urjATLCklGxcXkD/jNjmreKM/ftjnzSJw4s+JHHu7a1Kf8FGSslPNTHfwgAAIABJREFUn+/DLmrpZziEsW8EmXf1IgJRkjVAhhBiEHAAuAaY3XKAEGIs8CIwXUpZfOwpFHqHHWNKSrve6sd/QiNc9AzEDYL//FG7hZ+1sEvZXHtoveoCs94aOROlDdWQ9ykcWAuFG7XViE2bfAg9JGVqnh59R0G/U6DPyIB+ZwNHJWCy6Nm++lC3iLrf7aZ+40biZs/ufHAjBdsqWPneDvRGHeden0XGaX0AbYXpgd/cTe333+M444xQhcyB7Ycp3ldN1qEVOMaODtl1FB3TqahLKb1CiHnAUrSWxleklFuEEA8Da6WUS4Cn0Yqv7zX6g+yXUl4SwrgjEnNGRpttjV1GCDjzTohNhUW/hJfPg2vfC5p5UNOqUoOw9uzdj/x+2LdSW/a99SNtGbkpSusQGjO7Mfs+RRP0Lt7NGEx60scmsWt9CWfP8mEwhXYRmGvzFqTbjS17XMDv2bi8AGuUkdg+Nr741xbKC2sZf9EgHOedhz4mhsoPPgyZqLtdXr5ZuA2bw0Dy9i+wzvxtSK6j6JyA7vmllJ8Cnx712oMtvj8vyHH1SsxDh1KzciXS7UaYgui8OPLnWsfFwlnwr/PhmoWQ1tFcdmAk2E1YjXp00tIz9ymt2HvEIvXwfm0p+SlXwZhrISVb21EqiAwd35e87w+xd1MZQ8YlB/XcR1O3bi0A1lM7WXTUSGVJPXs3lTJu+gBOmzGIFQu2sfbTvZQX1nLejcOJDqHJl5SSFQu2UVFUx3mTPPg/9mAdOzao11AETgSYG/cezEOHgtdLw569wT952unw//6jLbF+7WLYsuiETymEwBlnxe8z95xM3V2rZeSvXgTPjoYVT2qrEWe+DL/dDhc/C6njgy7oACnD4rBFm9i++lDQz3009WvXYUpPP8YrqD02ryhACMHIyU70Bh1Tr8vkjCuGsCenhA//Zx2Gn12KdLup/PiToMe6deVBtq8uYvxFg4gt+AlhNmMZNizo14l0pJSUVDfw4+4yDlW6QnYdZRPQjbTsgLEMGxr8CyQMhlv+A2/P1vY63bUcJv3qhFbrOeOsbPIYw7tPqZSw/3tY/5bma+2u0VY/nnM/nHKNVn7qBnQ6QUZ2HzZ9U4Cr1oPFHoKl7Gjb09WtX0/0tMCcOd0uL1u/K2TwqUk44rSFS0IIxpyX9v/bO/O4OKtzj38Pwwyzse8QCJBAQvadbJpF69bExDWxtUartVZTra1Vu6Ra29vF2qtVr+aaaDXXNYkLMWrURKvRrEBCNgwQIIRAIKyBmQGGmXP/eAdCErYQYJjJ+/185jMzZ9555zkc5jfnfc5znofgKBOfrT7AhvXNTBg3j9p3+zbJ18niera+k0fcqBCmXJ1A0Wt70Y8d07dXoh6GtbmFwkoLhZUWCk623jdQUGmhvlGpH/unRaP50YyEfvl8VdQHEL+EBPD17bBgRp9hCoXb0uHzFZD5GmStUeKaZyzvPK1oF8SFGMk4rnVPndLaY5D9tpLvuqZQyV89erHiXomfMeB5qgFS0iLJ/uIYBXtOMmp2TL98RlNePs5TpzB0l+/FRe6ucpptLYybO+Sc14aOCeWGR6bw0Qv72Bl6AyMOrSGmN0m+OrLT1sKmVQfQm7V8745RyOYmGnNyCF122wWf29P4Ovck//v1EQpOWig7axYeE6gnKdzM4gmxJIWbSAwzMSY2sN9sUUV9ABE6HX6Jif0r6qAsBl7zD6WE1u5VsGsVHP4YYibBzJ8rqUY1PRv6IcEG7EU66u0n+9fmVpqt8N1GRcgLvgIkJFwCcx5Rcsy7KUd1K+Hx/gRFGsnddaLfRL3Vn26cMqXbY6WU7PuyhPB4f6KGdSwUIdEmbnpkCpte3EuOXIbjlQwuezL1guLtpZR8uSaH+qpGFv9yIgZ/nZJ8zH7x+dOLKi387PVMgow6ZiSFkhhmIincTGKYIuCGfl5UPxtV1AcYv5QUpcL6QGAOh3m/VUpyZb8F2/8H1t8BgfEw416YeKsrq17nxAUbkc5+9qlLqeS12fO6shbQdEqJE5/7qFKBJjih/z77PBFCkDw1kt0fFdJQ04g5uO9zzNsys/CNjEQb2/0O0JLDNdSUWbhsWWqXlan0Zi0LfzmZTQ+sIbchnqYX9nLFXWPR6XsnAfu+LOHInpPMuH4YMcOVhVdrVhYAhgkTenVOT6SpxcHyt7Lw1fiw9p4ZxAa5fy+HulA6wPglJ2MvLcXRMIALjzojTL0Tlu9WSmwFxsKmR+Hp0fD5Y0rkiOx4g++QYCPSqcfWopS06zOarUrq1a+ehOenKlE7+9fByAWwbCPcn62I+iAS9FZSpkaChLzdfb8lQ0qJNTMT4+RJPSofuP/LEvRmLcOndB+No9H4MPeHI0jJfZvig9Ws+2sGxw/XnLeNJwrr2PZuPgnjwpj4PSXdh2xpoXbdevRjxnRbmcmb+Psnhzlw/BRP3jhuUAg6qDP1AcevXW5146QBvkz10UDqAuV2bDdsfw62PQvfPqOU4wobDqHDITS57fEQcxw4/JA4sbXYMGq7Lo7QIVJC3TE4tst126ns6pQO5fX4mUq8/ahF3V45DAaCIo1EDPUnd/cJJl7RtzmM7MdLaTlxAsPk7uPTT1XaKNynhDH6ant2iW9MSyNRrCC0eTMHWhbywdN7SJkWycwbhveoOlGjxc6nqw5gCvI74+qgLn0D9uJiIl94oUd2eAObD5XzyreF3D4zgStHD57ds6qoDzBtop7nBlFvT9xUiFsD1YVK9r+qfKjMg6JvYd87bYcFAw+Zo3gOHZZND2M0xygFEjorKdba1tQAJS4BP7br9LZ8rRFiJ8PsByEuTUlS1Ye7YAeKlGlRfLMuj+oyCyHRfefnt52HP33/f1rDGHueqKs1yZf9uee58ZN72b/fTtZnRynaV8m0hUmMnRuLj6bjC3jplGx59RDWumau//XktugfabdTuXIl+lGjMM+b22NbPJnSWhsPrc9mdEwAv7lmpLvNOQNV1AcYbWwMPkZj/y+W9pSQRJj+szPbmi1KbvCqfKjKpyZzE3ASy3cbCbec5+V6ULxS2CMuDYZMVbbl93CRdjAzfEoE367PI3fXCab3YTZCa2YWPv7++A3vus6lvclBzrYyhk0MP2+/fmuSL8vGdNJ+vpwR06PY+k4u36zLI2dbKZfeMqLNT96ePZ8XU7S/ikuWpBCZENDWXrfhQ+zHjhH5wgs9chl5Oi0OJ794ey/NLU6eu2Uifr4DuxDaHZ7/7fIwhBBKuoDBIuodoTMpeVGixwGwqygYeBbLHR9B6CilMEP76u4t7au8Nyr3Gp0yIw+Idm9f+glToB9DRgaTt7uctGuT+kzMrJmZGCZNRGi6ForDO0/QZG1h3Lxzwxi7oy3J1/r1hN51J0ERRhYsH09hdiVb1+by/lNZjJgexczrh2MMUOLNS/Nq2JFewPDJEYyde/rKoG2WPnr0RTNLf3ZLHruKqnl6yXiSws+jYPUAoYq6G/BLSaH+s8+QUnrEzCbKHEhuA9Q31yux4VqXq+UiJ2VaFFtey6G88BRRSRced9xSU0PzkSMELlrU5XFSSvb/p4SwOHOnYYzdEfrTuym+bRknn3ueyId/jRCCpAnhxKWGkPlJEXs+L6Ywu5K0a5MYNjGcT1cfJCBMz7xbR57xP9s2S3/x4pilb8uv5Lkv87lx8hCum3j+P6gDgRr94gb8UlJw1NXRcnKAYr8vkNjAYAAqLKfcbMngImlCOBqtD7k7+yZtgM0VEthdkenjh2uoLrUwbl5cr4XUNG0aQUuWUP3qq9j2729r1/ppmL54GEtXTCMywZ+t7+Ty+ortNFlbuOruMWekiJZ2O5UvvqjM0ufO7ZUdnkRlQxMPvLOXpDATTyzq36pjF4Iq6m7AL7k1XUAfpeHtZ+KDlIXMktrzD3/zZnQGXxLGhpGfVYHD4bzg81kzMhE6HfqxY7s8bp8rjDF56oUlFYt46Ff4hoVR9rvfI5vPLKUYHGVi4f0TuPInYwiMMDDv1pGEDTkzMqluwwbsJSWELb/P62fpTqfkV2uzqbPZef4HkzDqBq+TQxV1N+A34nTBDE8gMVQR9bL6WjdbMvhImRaJrd5OSc6F/+BZMzPRjxuLTxd5U1rDGEfPjulxGGNnaPz9iXr8cZpyc6lcvfqc14UQDJ8cwdIVaYxIOzNkT5mlr0Q/ZsxFMUtftbWAr3JPsmLBKFKjA7p/gxtRRd0N+AYHowkPoynPM2bqyWFhAJQ3dFAk+SJn6JhQ/Iy+5O6+MBeM02ql8dAhjJO6jk/f/9VxJYxxTt/UG/WfP4+Aa66h6sWVNOXn9/h9F9MsPau4hn98epirx0Rxa9rgr62sirqb0CeneMxMPdxsBulDlVX1qZ+NxteHYZMiKNhbib3J0evz2Pbtg5aWLoti2Jsc5HxbStKE8w9j7IrI3/0WH5OJst+vQDq678MZs/Q5HZYj9hrqbHbuf2sPUYF6/nbDOI/4AVNF3U34JSfTlJ/foy+RuxFC4CP11DbWu9uUQUnKtEhamhwU7uv9wrc1IxOE6DIZVlsY4/y+jbrwDQ0l8re/wbZ3LzVvvNnt8XXp6RfFLL3a0sxD67I5UdfIs7dMJNDQP6mW+5rB6+33cvxSUpBNTdiPHUOXkOBuc7pF62PgVOMgKZQxyIgZHoQ52I+8XeWkTO3ddnFrZgZ+I0ag8e84TUL7MMboXoYxdkXAwoXUbdxIxdNPY54/H92Qjt07bbP0sWMH1Sy9zmZn5VdHCDP7cXlqBENDe7fL1+5w8tXhk6zPLGHLd+XYHZLffz+VSfHBfWxx/6GKuptoTRfQmJvrEaKu1xipcSX18ubZWW8QruIZ2VuOYWtoxmA+vwIR0m7Hlr2PoOuu6/SY1jDG+beN7Je/vxCC6Mcfp2DBQk489hhxq1d1+Dl16enYjx8n6g8rBs3/QWGlhTtf201hpQUp4U8bD5EcYeay1EguT41gYnwwmm7SDB8+Uc/6zGO8v6eUyoYmwsw6ls1I4IbJQwb9wujZqKLuJvyGDwMhlLDGK65wtzndYtKaqBKNVFmaCTN3n/jpYiMlLZI9nxdzJOvkeeViAWj87juk1dqlP/10GGPkhZraKdqYGMIf+hXlT/yJug/SCbpu8Rmvt5+lmy69tN/sOB++yavkvjez8BHw1k+mEx2oZ3NOBVtyylm9tYCVXx0hxKRj7ohwLk+N5NKUcMx+iuzVWpvZkF3K+swS9pXU4esjmD8ygpumxDF3RDjaTnLgDHZUUXcTPgYD2vg4j1ks9deZET4nKKmxqaLeAaGxZoKjTeTuOnHeom7NyATA0EHkS2GlhZc+Pkxsdi2p82MvOIyxO4KXLuXURx9T/re/Yb5kNr6uyCeA2g8+GFSz9DXbi/jjh4cYFm7i5WVTiQtRMojeOTuRO2cnUmez83XuSTbnlLMlp4L3so6j1QimJ4Vi9vNlS04FzQ4nI6P8WbFgFIsnxBDqBf/bqqi7EX1KiseENYYYAsDnKCU1VibE9W01em9ACEHKtEh2phdwqspGQGjP0yjYsjLRxsWhjTy9mSivvJ7nv8znw72lLLTqkPjw1JHjjKiJY0hwL9If9xDh40P0n/5E4eLFnPjzfzHkmacBkM3NVK3830ExS7c7nPzxw4O8vqOYy0ZG8MzSCfjrz13EDDRoWTg+hoXjY2hxOMk4WsMWl8DX2ez8IC2em6YMYXRM/5WWcweqqLsRv+QU6rd8gbOxER9931fQ6UvCTAEInyY2HyrnytFRHntp2p8MnRDGzvQC3lr3HTMXJDEqJqBbX65SFCMLs0soc8pO8fwX+Xx8oAyDVsPyiHD8chuImxNNyeFibl65nf+7K41hF5hIyumUHKuxUm1pptZqp9rSTI3V9djazLBZi5m5aS2/fiCBbdFj+Un9ftIGwSy9xtLMvW9ksb2gip/OSeLhK0d2+zcG8NX4MD0plOlJofzu+xden3Uwo4q6G/FLSQank6YjRzCMHry5JAAC/czodHY+2FvK8Vobz90yiajAwf1DNBA0tzj5Jv8kG7PL+PxQOQs1Puj3VXFdYSlmk5YZSaHMGh7GrOFhJIQazxHE5sJCHNXV1CSl8viaDD47VI7Zz5d75w5jvsHMjjdzGTkzmvlLRzKsLIbbXt7FzSu3s+bOab2eYeZXNPCrddlkHzt3h7DGRxBk0BKSdCmJB7Zz8zdvYf/JGBLWriU3JJ4PrRH81GZ3S3hffkU9d76WQVltI/+8aTw3TB6cCbXcjSrqbqR9wYzBLupmnZkWaePpJeP43fsHuebZrTy9ZAJzUsLdbdqAY3c42Xakio3ZpXx68ASnGlsI0Pty1ZgoZphMFH9YzCPGEPKS9XxTVM0nB5TdprFBBmYNV0R+5rAwwv39OPzZ12iBu7Md1IdX8cBlyfx4ViKW4w2kP7OX2BFBzP3BCIQQjI4JZO09M7h19U6WvrSDV++YyuShPS8w4nRKXvm2kH98ehiDTsNjC0cxNNRIsFHXdvPX+7YVpLZdE0XRzUtY/t6T2K3VZCy+ixe+KuD1ncXcM3cYd8xMHLCiyl8eruD+N/fgp/XhrbunM3mo54QYDjSqqLsRXXw8QqfziMReJl8l7veKMSGMjZ3NfW9kseyVXdw3bxgPXp6Cr5e7YxxOyY6CKjbuK2PTgTJqrHbMfr5cMSqSBeOjmT08HJ2v8jfIC/dn878PMUXry+M/n015Uwvf5FeyLb+STw+WszajBFBEfumWzUzVm7nl+lncNiuRAL2W2nIrH6/cT2C4gavuHovG9/Tfdli4mXUuYb919S5W3TaF2clhHdrcnqNVFn69bh+7iqq5PDWCv1w/lgj/rq+0DKNHE/rjO6hatRr9uHH8asXtXFNWz1OfHebJTYf597dF3D9/OEumxrf1va+RUvLyN4X85eMcRkYFsGrZlEFTC3SwIvq0mPB5MGXKFJmRkeGWzx5MFFx/Pb4hocSvXuVuU7pkXe46ntj+BJtv3EykKRJbs4PHNxzknYxjpCWG8OwtE4kM8D53TJ3Vzus7j/LatiIq6psw6jRclhrJgnHRzEkJR99JNErR/ko2vXSAgDADix6YgClIiapwOCWHSk/xTX4lWcU1/OylhwgZP5aE/3kOgMYGO+v/nkGTrYUbH5lCYHjHAlZR38htL++i4KSF538wkSs6qZEppeSNncX85eMcNELw2LWjuWFSbI/94s7GRsr/8leCbr4Zw5jTV5O7i6r5x6bD7CqqJj7EyIPfS+ba8bHd+rcdTskpm50aq+LDr7G0e2y1U+tqq7Y2U2ttptpip7KhiStHR/L0kgmDOjviQCGEyJRSdlrvUBV1N1P6yCNYtu8g+euv3G1Kl3xS+AkPf/0w6YvSSQpKamt/N7OE339wAJOfhmeWTOzRrLE/OdVoZ3dhNUOCjaREmnu9qFdcZeWVbwtZm3EMa7ODS5LDWDo1nvkjI3rscjh+uIaPXtiHwV/LtQ9MPEeg7eXl5M+ZS8SjjxB6++047E7S/7WHiqJ6Fj04sdudo7XWZm7/9272H6/jqZvGnVO0obTWxiPv7mNrXiWXJIfx9xvGEdOHs1wpJf/JPck/Nh3mUNkpRkT6s2RqHDa7o02Qa61nLsDW2ex0Jjm+PoJgk45go5Ygo44Qo45gk5ZRMYH8cFp8m1voYqc7UVd/9tyMX0oKdekbcNTWogkavKGCJq3ifrHYLWe03zB5COOGBHLvG1n86JWd/Hx+Mg9cltyjiIS+orjKqsQif1fOzoJqWpyKagwJNnB5aiSXpUaQlhjaIxfBnuIaVm8t5JMDZWh8BNeOj+WuSxJ7taswdkQwi34xkQ+f38v7T2Vy7QMTCYk5vX3dlqnEpxsnT0FKyRev51CWX8cVd43uUSqAIKOO1+9K4+41GTz4TjYNjS38aEYCUkrWZ5bwxIeHcEjJnxeP4Ydp8X0etSKEYN6ICOYkh/PR/jL++/Ncnth4CACDVtMmzsEmLTFBBpff/nRbqx8/xKQjyKjF7Oc7KOLfPR1V1N1M+8VS49Spbramc1pFvcF+bv6X5Eh/0pfPYsUHB3l2Sx4ZRdU8s3RCtz7b3uJwSvYeq2nbOZhbrtg0PMLMnZckMic5nKPVVrbklPPWrmJe3VaE2c+XS1PCuDw1knkjIgg2nd7K73RKNueUs2prAbuLavDX+3L3pcO4fWbCBUf4RCYGcN0vJ7HhX3t5/59ZLLx/PBFDlR8Ia0YmwmhEnzqS3R8VkbtTqXeaPKXnu0bNfr68cvtUlr+ZxYr0g1TUN5FTdorNORVMSwjhqZvGEx/af3HtAD4+goXjY7h6TBRVlmYCDdpO3VIq/Y8q6m6mtQpS7fsf4BMYiF9y8qCcrbSKutVu7fB1o86Xf948nrSkEP6QfoC0v2zBoNVg0GrQazXotT7oXc8NOg1+vsq9oV273nUzaH0w6No/V+6rGprYnFPBl4crqLY04+sjmJYYwpKp8eckcZoJ3DItHluzg2/zK9nynbLp5OP9J/ARMHloMJelRmLSafj3t0UUVFqIDTLwhwWjuHlqXNtW8r4gNNbM9b+eRPoze/ng6T0suG880Un+WDMyME4YT25mJbs3FjJyehSTrx563ufXazW8eOtkHlqXzXNf5OPn68Pvv5/Kj2clDqjLwlfj45XrKp5Gj3zqQoirgH8BGmC1lPJvZ73uB6wBJgNVwBIpZVFX51R96gpSSo7e+qO2S3FNWBim6dMxzZiOacYMtDExbrZQ4Vj9Ma557xr+POvPLBredWHkvPJ6PswuxdrswGZ30Gh30mhvfXxWW7ODxhblvqml+5JwgQYtc0eEc1lqJHNSws8rXtrplBworWNzTgWbD5VzqEzJDz9+SCA/uTSJq0ZH9VsUj6O+nspte/n0owYsTRrG5b5KSNkeuOtR/lMUT3RSIAvvn3BGpMv54nQqbpfJCcEXvDlJZfBywQulQggNkAt8DygBdgO3SCkPtTvmXmCclPIeIcRS4Dop5ZKuzquK+pnYjx/HsmMHlu07sOzYgaOyEgDd0KEYZ0zHNGMmprRpbvO7VzdWM+edOTw67VF+mPrDfvkMp1PS1OI8Q/wVsXdga3Zi0PkwfkhQnwlvaa2NGmszo6ID+vTqSEqJ/ehRrHv2YtuzB9uePUpVISlp9gtg35QHqdeGMSnVzsFiE3qzjhsenoze5Bn5ulXcS1+I+gzgcSnlla7nvwGQUv613TGfuo7ZLoTwBU4A4bKLk6ui3jlSSpry8rDu2IFl23asu3fjtFhACHTx8aAdeK+ZlJKCugI0QoNGqP7SrjBZHJgtSvETm96H4ngDxfF6jg41UBKnx6ExMCHzWoJrY2jW2tg1Yy02o1oq8GLinvH3cFXiVb16b19Ev8QCx9o9LwHSOjtGStkihKgDQoHKs4y5G7gbID5+8Nf6cxdCCPQpKehTUgi57TYl3/b+A1h2bFc2KrkpDFVXK86JflE5l1q9L+WJQVQMC6Imygzt/Nqt+RtrLs+GA1Ys8WVEh4SifF1ULhYCdP2Xo70not7RdenZqtKTY5BSvgS8BMpMvQefrQIIrRbjpIkYJ3Ve6mwgUDNt9DGXu9sAFW+kJ87JEiCu3fMhQGlnx7jcL4FAdV8YqKKioqLSc3oi6ruBZCFEohBCBywFNpx1zAZgmevxjcAXXfnTVVRUVFT6h27dLy4f+XLgU5SQxleklAeFEE8AGVLKDcDLwP8JIfJRZuhL+9NoFRUVFZWO6VEYhZTyY+Djs9r+0O5xI3BT35qmoqKionK+eHe+VBUVFZWLDFXUVVRUVLwIVdRVVFRUvAhV1FVUVFS8CLcVyRBCnASO9vLtYZy1W9UL8LY+eVt/wPv65G39Ae/rU0f9GSql7LQ4sNtE/UIQQmR0lfvAE/G2Pnlbf8D7+uRt/QHv61Nv+qO6X1RUVFS8CFXUVVRUVLwITxX1l9xtQD/gbX3ytv6A9/XJ2/oD3ten8+6PR/rUVVRUVFQ6xlNn6ioqKioqHaCKuoqKiooX4XGiLoS4SghxWAiRL4R41N32XChCiCIhxH4hxF4hhEfW9xNCvCKEqBBCHGjXFiKE+FwIkee6D3anjedDJ/15XAhx3DVOe4UQ17jTxvNFCBEnhPhSCJEjhDgohHjA1e6R49RFfzx2nIQQeiHELiFEtqtPf3S1JwohdrrG6B1XCvTOz+NJPvWeFMH2NIQQRcAUKaXHbpgQQlwKNABrpJRjXG1PAtVSyr+5fnyDpZSPuNPOntJJfx4HGqSUT7nTtt4ihIgGoqWUWUIIfyATWAzcjgeOUxf9uRkPHSehVD83SSkbhBBa4BvgAeCXwHtSyreFECuBbCnli52dx9Nm6tOAfCllgZSyGXgbWORmmy56pJRfc26lq0XAa67Hr6F84TyCTvrj0Ugpy6SUWa7H9UAOSslUjxynLvrjsUiFBtdTresmgfnAeld7t2PkaaLeURFsjx5IlEH7TAiR6SrM7S1ESinLQPkCAhFutqcvWC6E2Odyz3iEm6IjhBAJwERgJ14wTmf1Bzx4nIQQGiHEXqAC+Bw4AtRKKVtch3SreZ4m6j0qcO1hzJJSTgKuBu5zXfqrDD5eBIYBE4Ay4J/uNad3CCHMwLvAL6SUp9xtz4XSQX88epyklA4p5QSUWtDTgNSODuvqHJ4m6j0pgu1RSClLXfcVwPsoA+kNlLv8nq3+zwo323NBSCnLXV84J7AKDxwnl5/2XeANKeV7rmaPHaeO+uMN4wQgpawF/gNMB4KEEK1V6rrVPE8T9Z4UwfYYhBAm1yIPQggTcAVwoOt3eQzti5EvA9LdaMsF0yp8Lq7Dw8bJtQj3MpAjpfzvdi955Dh11h9PHichRLgQIsj12ABcjrJW8CVwo+uwbsfIo6JfAFwhSs9wugj2f7nZpF4jhEhCmZ2DUi/2TU/sjxDiLWAuSprQcuAx4ANgLRAPFAM3SSk9YvGxk/7MRbmkl0DVSRJ7AAAAhklEQVQR8NNWX7QnIISYDWwF9gNOV/NvUfzQHjdOXfTnFjx0nIQQ41AWQjUoE+61UsonXDrxNhAC7AFulVI2dXoeTxN1FRUVFZXO8TT3i4qKiopKF6iirqKiouJFqKKuoqKi4kWooq6ioqLiRaiirqKiouJFqKKuoqKi4kWooq6ioqLiRfw/JkZZzwLfWj8AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(treatment_logs[:,:,:].mean(axis=1).transpose())\n",
    "plt.legend(algo_list)\n",
    "#plt.yscale(\"log\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "def compute_is_weights(dataset, q_policy):\n",
    "    n = dataset[\"observations\"].shape[0]\n",
    "    horizon = dataset[\"observations\"].shape[1]\n",
    "    \n",
    "    rewards = dataset[\"rewards\"][:,:,0]\n",
    "    \n",
    "    pies = np.zeros((n, horizon))\n",
    "    pibs = np.zeros((n, horizon))\n",
    "    length = horizon*np.ones((n))\n",
    "    \n",
    "    for idx in range(n):\n",
    "        for t in range(horizon):\n",
    "            pie = q_policy.get_prob(dataset[\"observations\"][idx,t,:], dataset[\"pibs\"][idx,t,:], \n",
    "                                    dataset[\"nn_action_dist\"][idx,t,:])\n",
    "            a = dataset[\"actions\"][idx,t,0]\n",
    "            \n",
    "            pies[idx,t] = pie[a]\n",
    "            pibs[idx,t] = dataset[\"pibs\"][idx,t,a]\n",
    "    return pibs, pies, rewards, length.astype(int)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "def wis_weights_on_neg_rewards(pibs, pies, rewards, length, weighted=True):\n",
    "    n = pibs.shape[0]\n",
    "    weights = np.ones((n,30))\n",
    "    \n",
    "    for i in range(n):\n",
    "        last = 1\n",
    "        for t in range(int(length[i])):\n",
    "            weights[i,t] = last*(pies[i,t]/pibs[i,t])    \n",
    "            last = weights[i,t]\n",
    "        weights[i,length[i]:] = weights[i,length[i]-1]\n",
    "    weights = np.clip(weights, 0, 1e3)\n",
    "    if weighted:\n",
    "        weights_norm = weights.sum(axis=0)\n",
    "    else:\n",
    "        weights_norm = weights.shape[0]\n",
    "    weights /= weights_norm\n",
    "    reward_filter = (rewards.sum(axis=-1) < -2).astype(float)\n",
    "    \n",
    "    return (weights[:,-1] * reward_filter).sum(axis=0), reward_filter, weights[:,-1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(f\"data/cancer_pcv1_val_episodes\", 'rb') as f:\n",
    "        train_dataset = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_keys(['observations', 'actions', 'rewards', 'sparse_rewards', 'dense_rewards', 'not_done', 'pibs', 'nn_action_dist'])"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_dataset.keys()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "#plt.hist(train_dataset['rewards'][:,:,0].sum(axis=-1))\n",
    "reward_filter = (train_dataset['rewards'][:,:,0].sum(axis=-1) < -2).astype(float)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.95"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reward_filter.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "New wis: 2.2 ESS: 2.01 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv1', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 2.34 ESS: 1.89 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv1', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 3}\n",
      "New wis: 2.64 ESS: 1.34 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv1', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 5}\n",
      "New wis: 2.96 ESS: 1.26 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv1', 'thresh': 0.5, 'var_coeff': 0.0, 'seed': 5}\n",
      "Neg-reward weights: cancer_mdp_pcv1 PG_mask 1.4842282765434383e-06 , neg rewards: 0.937\n",
      "New wis: 2.32 ESS: 2.31 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 3.0 ESS: 1.18 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 5}\n",
      "New wis: 3.01 ESS: 1.16 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'var_coeff': 0.1, 'seed': 5}\n",
      "New wis: 3.01 ESS: 1.15 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'var_coeff': 1.0, 'seed': 5}\n",
      "Neg-reward weights: cancer_mdp_pcv1 PG_nomask 8.30169215909906e-07 , neg rewards: 0.937\n",
      "New wis: 3.06 ESS: 1.01 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 3.07 ESS: 1.02 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 2}\n",
      "New wis: 3.08 ESS: 1.01 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 3}\n",
      "New wis: 3.08 ESS: 1.0 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 5}\n",
      "Neg-reward weights: cancer_mdp_pcv1 BCQ 0.001521102217871536 , neg rewards: 0.937\n",
      "New wis: 3.07 ESS: 1.01 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'beta': 1.0, 'seed': 1}\n",
      "New wis: 3.07 ESS: 1.02 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'beta': 1.0, 'seed': 2}\n",
      "New wis: 3.08 ESS: 1.01 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'beta': 1.0, 'seed': 3}\n",
      "New wis: 3.08 ESS: 1.0 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv1', 'thresh': 0.0, 'beta': 1.0, 'seed': 4}\n",
      "Neg-reward weights: cancer_mdp_pcv1 PQL 0.00010479063097027527 , neg rewards: 0.937\n",
      "New wis: -1.43 ESS: 14.28 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv2', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: -1.41 ESS: 4.6 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv2', 'thresh': 0.1, 'var_coeff': 0.0, 'seed': 5}\n",
      "New wis: 0.6 ESS: 1.61 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv2', 'thresh': 0.5, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 0.74 ESS: 2.11 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv2', 'thresh': 0.5, 'var_coeff': 0.0, 'seed': 5}\n",
      "New wis: 0.74 ESS: 2.1 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv2', 'thresh': 0.5, 'var_coeff': 0.1, 'seed': 5}\n",
      "New wis: 0.76 ESS: 2.06 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv2', 'thresh': 0.5, 'var_coeff': 1.0, 'seed': 5}\n",
      "Neg-reward weights: cancer_mdp_pcv2 PG_mask 0.010296521770968282 , neg rewards: 0.933\n",
      "New wis: 0.6 ESS: 1.58 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv2', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 0.76 ESS: 2.18 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv2', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 5}\n",
      "Neg-reward weights: cancer_mdp_pcv2 PG_nomask 0.00023147101471134913 , neg rewards: 0.933\n",
      "New wis: 0.58 ESS: 1.0 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv2', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 1.84 ESS: 1.0 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv2', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 2}\n",
      "Neg-reward weights: cancer_mdp_pcv2 BCQ 0.03792114100728035 , neg rewards: 0.933\n",
      "New wis: 2.35 ESS: 1.02 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv2', 'thresh': 0.0, 'beta': 1.0, 'seed': 1}\n",
      "New wis: 2.35 ESS: 1.02 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv2', 'thresh': 0.0, 'beta': 2.0, 'seed': 1}\n",
      "New wis: 2.35 ESS: 1.02 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv2', 'thresh': 0.0, 'beta': 5.0, 'seed': 1}\n",
      "Neg-reward weights: cancer_mdp_pcv2 PQL 0.002785796807156326 , neg rewards: 0.933\n",
      "New wis: 1.49 ESS: 2.45 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv3', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 1}\n",
      "Neg-reward weights: cancer_mdp_pcv3 PG_mask 0.14952891836259927 , neg rewards: 0.942\n",
      "New wis: 0.28 ESS: 3.72 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv3', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 0.96 ESS: 3.11 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv3', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 5}\n",
      "Neg-reward weights: cancer_mdp_pcv3 PG_nomask 0.00018154453464282866 , neg rewards: 0.942\n",
      "New wis: 2.54 ESS: 1.12 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv3', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 2.63 ESS: 1.11 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv3', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 3}\n",
      "Neg-reward weights: cancer_mdp_pcv3 BCQ 0.14554392176219882 , neg rewards: 0.942\n",
      "New wis: 2.5 ESS: 1.21 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv3', 'thresh': 0.0, 'beta': 1.0, 'seed': 1}\n",
      "New wis: 2.5 ESS: 1.21 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv3', 'thresh': 0.0, 'beta': 2.0, 'seed': 1}\n",
      "New wis: 2.58 ESS: 1.11 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv3', 'thresh': 0.0, 'beta': 5.0, 'seed': 2}\n",
      "New wis: 2.65 ESS: 1.05 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv3', 'thresh': 0.0, 'beta': 5.0, 'seed': 4}\n",
      "Neg-reward weights: cancer_mdp_pcv3 PQL 0.14474630437637603 , neg rewards: 0.942\n",
      "New wis: -1.08 ESS: 5.87 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv4', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: -0.77 ESS: 3.04 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv4', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 5}\n",
      "New wis: -0.12 ESS: 5.53 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv4', 'thresh': 0.1, 'var_coeff': 0.0, 'seed': 1}\n",
      "Neg-reward weights: cancer_mdp_pcv4 PG_mask 0.3286410939730349 , neg rewards: 0.95\n",
      "New wis: -0.48 ESS: 9.18 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: -0.35 ESS: 4.45 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 5}\n",
      "New wis: -0.35 ESS: 4.69 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'var_coeff': 0.1, 'seed': 5}\n",
      "Neg-reward weights: cancer_mdp_pcv4 PG_nomask 2.3730088847458087e-06 , neg rewards: 0.95\n",
      "New wis: 0.31 ESS: 2.19 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 0.42 ESS: 1.39 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 2}\n",
      "New wis: 0.53 ESS: 1.05 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 3}\n",
      "Neg-reward weights: cancer_mdp_pcv4 BCQ 0.22893813553787357 , neg rewards: 0.95\n",
      "New wis: 0.19 ESS: 2.29 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'beta': 1.0, 'seed': 1}\n",
      "New wis: 0.41 ESS: 1.77 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'beta': 1.0, 'seed': 2}\n",
      "New wis: 0.42 ESS: 1.13 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'beta': 1.0, 'seed': 3}\n",
      "New wis: 0.61 ESS: 1.0 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv4', 'thresh': 0.0, 'beta': 1.0, 'seed': 5}\n",
      "Neg-reward weights: cancer_mdp_pcv4 PQL 0.47309105431507 , neg rewards: 0.95\n",
      "New wis: -2.19 ESS: 6.86 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv5', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: -1.62 ESS: 1.96 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv5', 'thresh': 0.05, 'var_coeff': 0.0, 'seed': 4}\n",
      "New wis: 1.43 ESS: 1.12 at: {'name': 'PG_mask', 'env': 'cancer_mdp_pcv5', 'thresh': 0.1, 'var_coeff': 0.0, 'seed': 1}\n",
      "Neg-reward weights: cancer_mdp_pcv5 PG_mask 0.001318504355944473 , neg rewards: 0.933\n",
      "New wis: 0.8 ESS: 1.73 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 1.01 ESS: 1.08 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 5}\n",
      "New wis: 1.01 ESS: 1.07 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'var_coeff': 0.1, 'seed': 5}\n",
      "New wis: 1.05 ESS: 1.03 at: {'name': 'PG_nomask', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'var_coeff': 1.0, 'seed': 5}\n",
      "Neg-reward weights: cancer_mdp_pcv5 PG_nomask 0.0016739346785781366 , neg rewards: 0.933\n",
      "New wis: 1.07 ESS: 1.98 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 1}\n",
      "New wis: 1.18 ESS: 1.43 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 3}\n",
      "New wis: 1.52 ESS: 1.01 at: {'name': 'BCQ', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'var_coeff': 0.0, 'seed': 5}\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Neg-reward weights: cancer_mdp_pcv5 BCQ 0.005687385065999652 , neg rewards: 0.933\n",
      "New wis: 0.75 ESS: 1.08 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'beta': 1.0, 'seed': 1}\n",
      "New wis: 1.52 ESS: 1.01 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'beta': 1.0, 'seed': 2}\n",
      "New wis: 1.52 ESS: 1.01 at: {'name': 'PQL', 'env': 'cancer_mdp_pcv5', 'thresh': 0.0, 'beta': 2.0, 'seed': 2}\n",
      "Neg-reward weights: cancer_mdp_pcv5 PQL 0.0004560516525234003 , neg rewards: 0.933\n"
     ]
    }
   ],
   "source": [
    "env_list = [\"cancer_mdp_pcv1\", \"cancer_mdp_pcv2\", \"cancer_mdp_pcv3\", \"cancer_mdp_pcv4\", \"cancer_mdp_pcv5\"]\n",
    "env = CancerSim(dose_penalty=1.0, therapy_type=\"PCV\", env_seed=None, max_steps=30, state_dim=5)\n",
    "\n",
    "#env_list = [\"cancer_pcv1\", \"cancer_pcv2\", \"cancer_pcv3\", \"cancer_pcv4\", \"cancer_pcv5\"]\n",
    "#env = CancerSim(dose_penalty=1.0, therapy_type=\"PCV\", env_seed=None, max_steps=30, state_dim=3)\n",
    "\n",
    "algo_list = [\"PG_mask\", \"PG_nomask\", \"BCQ\", \"PQL\"]\n",
    "\n",
    "weights_results = np.zeros((len(algo_list), len(env_list)))\n",
    "\n",
    "for j,env_name in enumerate(env_list):\n",
    "    with open(f\"data/{env_name}_train_episodes\", 'rb') as f:\n",
    "        train_dataset = pickle.load(f)\n",
    "    with open(f\"data/{env_name}_val_episodes\", 'rb') as f:\n",
    "        val_dataset = pickle.load(f)\n",
    "    trees = construct_nearest_neighbor_trees(train_dataset[\"observations\"],train_dataset[\"actions\"])\n",
    "    for i,algo in enumerate(algo_list):\n",
    "        param, val, train = get_best_hp(env_name, algo)\n",
    "        model, _, _, _ = get_model_from_param(param)\n",
    "        pibs, pies, rewards, length = compute_is_weights(train_dataset, model)\n",
    "        weights_sum, reward_filter, weights = wis_weights_on_neg_rewards(pibs, pies, rewards, length)\n",
    "        print(\"Neg-reward weights:\", env_name, algo, weights_sum, \", neg rewards:\", reward_filter.mean())\n",
    "        weights_results[i,j] = weights_sum"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "val_weights = weights_results.mean(axis=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.01990315, 0.0371038 , 0.00487675, 0.0151123 ])"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "val_weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Non MDP"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.05499454, 0.0056951 , 0.0003608 , 0.00514889])"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_weights = weights_results.mean(axis=-1)\n",
    "train_weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "# MDP"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.0979573 , 0.00041803, 0.08392234, 0.1242368 ])"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_weights = weights_results.mean(axis=-1)\n",
    "train_weights"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Val-Test correlation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_all_checkpoints(env, alg, verbose=True):\n",
    "    #var_coeff = 0.0\n",
    "    var_coeff_list = [0.0, 0.1, 1.0]\n",
    "    seeds = [1,2,3,4,5]\n",
    "    if alg == \"PG_mask\":\n",
    "        threshold_list = [0.05, 0.1, 0.5]\n",
    "    elif alg == \"PG_nomask\":\n",
    "        threshold_list = [0.0]\n",
    "    elif alg == \"PG_pib_mask\":\n",
    "        with open(f\"data/{env_name}_train_episodes\", 'rb') as f:\n",
    "            train_dataset = pickle.load(f)\n",
    "        with open(f\"data/{env_name}_val_episodes\", 'rb') as f:\n",
    "            val_dataset = pickle.load(f)\n",
    "        return {\"name\": alg, \"env\": env, \"thresh\": 0.2, \"var_coeff\": 0, \"seed\": 0}, wis_fix_policy(val_dataset), wis_fix_policy(val_dataset)\n",
    "    elif alg == \"BCQ\":\n",
    "        var_coeff_list = [0.0]\n",
    "        threshold_list = [0.0, 0.25]\n",
    "    elif alg == \"PQL\":\n",
    "        var_coeff_list = [1.0, 2.0, 5.0]\n",
    "        threshold_list = [0.0, 0.25]\n",
    "    best_wis = -100\n",
    "    score_list = np.array([])\n",
    "    test_score_list = np.array([])\n",
    "    for var_coeff in var_coeff_list:\n",
    "        for threshold in threshold_list:\n",
    "            for seed in seeds:\n",
    "                if alg == \"PQL\":\n",
    "                    param = {\"name\": alg, \"env\": env, \"thresh\": threshold, \"beta\": var_coeff, \"seed\": seed}\n",
    "                else:\n",
    "                    param = {\"name\": alg, \"env\": env, \"thresh\": threshold, \"var_coeff\": var_coeff, \"seed\": seed}\n",
    "                model, score, train_score, test_score = get_model_from_param(param)\n",
    "        \n",
    "                score_list = np.append(score_list, score[:,0])\n",
    "                test_score_list = np.append(test_score_list, test_score)\n",
    "    return score_list, test_score_list\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "env_list = [\"cancer_mdp_pcv1\", \"cancer_mdp_pcv2\", \"cancer_mdp_pcv3\", \"cancer_mdp_pcv4\", \"cancer_mdp_pcv5\"]\n",
    "algo_list = [\"PG_mask\", \"PG_nomask\", \"BCQ\", \"PQL\"]\n",
    "for j,env_name in enumerate(env_list):\n",
    "    plt.figure()\n",
    "    for i,algo in enumerate(algo_list):\n",
    "        score, test_score = get_all_checkpoints(env_name, algo)\n",
    "        plt.scatter(score, test_score, label=algo)\n",
    "    plt.xlabel(\"Validation SNTIS\")\n",
    "    plt.ylabel(\"Online test value\")\n",
    "    plt.legend()\n",
    "    #plt.ylim([-20,3])\n",
    "    plt.title(env_name)\n",
    "    plt.savefig(\"val_test_env\"+str(j)+\".png\")\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}