examples/quickstart/rec/rec_synthetic_discrete_basic.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "# Quickstart Example with Synthetic Recommender Simulation and Dataset (Discrete)\n",
        "This notebook provides an example of conducting **Off-policy Evaluation (OPE) and Offline Reinforcement Learning (Offline RL) on a synthetic recommender system (REC) dataset**.\n",
        "\n",
        "This example consists of the following 5 steps: \\\n",
        "0. (Setup Synthetic Recommender Simulation Environment and Online RL)\n",
        "1. Synthetic Data Generation\n",
        "2. Offline Reinforcement Learning\n",
        "3. Off-Policy Evaluation and Evaluation of Offline RL methods\n",
        "4. Evaluation of OPE methods\n",
        "\n",
        "\\* This library uses [d3rlpy](https://github.com/takuseno/d3rlpy)'s algorithm implementations of online/offline rl policies and model-based evaluation.  \n",
        "\\* Also, our implementations of OPE are highly inspired by [Open Bandit Pipeline](https://github.com/st-tech/zr-obp)."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {},
      "outputs": [],
      "source": [
        "# import SCOPE-RL modules\n",
        "import scope_rl\n",
        "from recgym import RECEnv\n",
        "from scope_rl.dataset import SyntheticDataset\n",
        "from scope_rl.policy import OnlineHead\n",
        "from scope_rl.policy import EpsilonGreedyHead\n",
        "from scope_rl.ope.online import (\n",
        "    calc_on_policy_policy_value,\n",
        "    visualize_on_policy_policy_value,\n",
        ")\n",
        "\n",
        "# import d3rlpy algorithms\n",
        "from d3rlpy.algos import DiscreteRandomPolicyConfig\n",
        "from d3rlpy.algos import RandomPolicyConfig as ContinuousRandomPolicyConfig\n",
        "from d3rlpy.preprocessing import MinMaxObservationScaler, MinMaxActionScaler\n",
        "\n",
        "# import from other libraries\n",
        "import gym\n",
        "import torch"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {},
      "outputs": [],
      "source": [
        "import pickle\n",
        "from glob import glob\n",
        "from tqdm import tqdm\n",
        "\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "import seaborn as sns\n",
        "\n",
        "%matplotlib inline"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "0.1.2\n"
          ]
        }
      ],
      "source": [
        "# version\n",
        "print(scope_rl.__version__)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {},
      "outputs": [],
      "source": [
        "# set random state\n",
        "random_state = 12345"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {},
      "outputs": [],
      "source": [
        "device = \"cuda:0\" if torch.cuda.is_available() else \"cpu\""
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {},
      "outputs": [],
      "source": [
        "# log directory\n",
        "from pathlib import Path\n",
        "Path(\"logs/\").mkdir(exist_ok=True)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {},
      "outputs": [],
      "source": [
        "# ignore warnings\n",
        "import warnings\n",
        "warnings.simplefilter('ignore')"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 0. (Setup Synthetic Recommender Simulation Environment and Online RL)\n",
        "To begin with, we briefly describe how the recommender environment works for the preparation.\n",
        "\n",
        "#### RL setup for recommendation\n",
        "In recommendation, the objective of the RL agent is to maximize reward.\n",
        "\n",
        "We often formulate this recommendation problem as the following (Partially Observable) Markov Decision Process ((PO)MDP):\n",
        "- `state`: \n",
        "   - A vector representing user preference.  The preference changes over time in an episode by the actions presented by the RL agent.\n",
        "   - When the true state is unobservable, the agent uses observation instead of state.\n",
        "- `action`: Indicating which item to present to the user.\n",
        "- `reward`: User engagement signal. Either binary or continuous.\n",
        "\n",
        "\n",
        "For more about the environmental configuration , please refer to [examples/quickstart/rec/rec_synthetic_customize_env.ipynb](https://github.com/hakuhodo-technologies/scope-rl/blob/main/examples/quickstart/rec/rec_synthetic_customize_env.ipynb).\n",
        "\n",
        "Let's see how it works!"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {},
      "outputs": [],
      "source": [
        "# setup environment\n",
        "env = gym.make('RECEnv-v0')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {},
      "outputs": [],
      "source": [
        "# define a random agent\n",
        "agent = OnlineHead(\n",
        "    DiscreteRandomPolicyConfig().create(device=device),\n",
        "    name=\"random\",\n",
        ")\n",
        "agent.build_with_env(env)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "metadata": {},
      "outputs": [],
      "source": [
        "# interact agent with the environment\n",
        "# only 6 lines are needed for RL interaction\n",
        "for episode in range(10):\n",
        "    obs, info = env.reset()\n",
        "    done = False\n",
        "\n",
        "    while not done:\n",
        "        action = agent.predict_online(obs)\n",
        "        obs, reward, done, truncated, info = env.step(action)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "(5,)\n",
            "[ 0.13726447  0.0194582   0.10000844 -0.04762143  0.09446818]\n"
          ]
        }
      ],
      "source": [
        "# state \n",
        "print(obs.shape)\n",
        "print(obs)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 14,
      "metadata": {},
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "# let's visualize the transition of reward\n",
        "obs, info = env.reset()\n",
        "done = False\n",
        "reward_list = []\n",
        "\n",
        "while not done:\n",
        "    action = agent.sample_action_online(obs)\n",
        "\n",
        "    obs, reward, done, truncated, info = env.step(action)\n",
        "    reward_list.append(reward)\n",
        "\n",
        "# plot\n",
        "fig = plt.figure()\n",
        "ax1 = fig.add_subplot(111)\n",
        "ax1.plot(reward_list[:-1], label='reward', color='tab:orange')\n",
        "ax1.set_xlabel('timestep')\n",
        "ax1.set_ylabel('reward')\n",
        "ax1.legend(loc='upper left')\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "For more about the environmental configurations and customization, please refer to [examples/quickstart/rec/rec_synthetic_customize_env.ipynb](https://github.com/hakuhodo-technologies/scope-rl/blob/main/examples/quickstart/rec/rec_synthetic_customize_env.ipynb)."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "#### Online RL and comparison of the policies\n",
        "We now learn an online RL policy and compare it to the random policy, using on-policy policy estimation."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 15,
      "metadata": {},
      "outputs": [],
      "source": [
        "# standardized environment for discrete action\n",
        "env = RECEnv(random_state=random_state)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {},
      "outputs": [],
      "source": [
        "from d3rlpy.algos import DoubleDQNConfig\n",
        "from d3rlpy.models.encoders import VectorEncoderFactory\n",
        "from d3rlpy.models.q_functions import MeanQFunctionFactory\n",
        "from d3rlpy.dataset import create_fifo_replay_buffer\n",
        "from d3rlpy.algos import LinearDecayEpsilonGreedy\n",
        "\n",
        "# model\n",
        "ddqn = DoubleDQNConfig(\n",
        "    encoder_factory=VectorEncoderFactory(hidden_units=[30, 30]),\n",
        "    q_func_factory=MeanQFunctionFactory(),\n",
        "    target_update_interval=100,\n",
        ").create(device=device)\n",
        "\n",
        "# replay buffer\n",
        "buffer = create_fifo_replay_buffer(\n",
        "    limit=10000,\n",
        "    env=env,\n",
        ")\n",
        "\n",
        "# explorers\n",
        "explorer = LinearDecayEpsilonGreedy(\n",
        "    start_epsilon=1.0,\n",
        "    end_epsilon=0.3,\n",
        "    duration=1000,\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "# start training\n",
        "# skip if there is a pre-trained model\n",
        "ddqn.fit_online(\n",
        "    env,\n",
        "    buffer,\n",
        "    explorer=explorer,\n",
        "    eval_env=env,\n",
        "    n_steps=50000,\n",
        "    n_steps_per_epoch=1000,\n",
        "    update_start_step=1000,\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 18,
      "metadata": {},
      "outputs": [],
      "source": [
        "# save model\n",
        "ddqn.save_model(\"d3rlpy_logs/ddqn.pt\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 19,
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-07-22 17:54:09 [warning  ] Parameters will be reinitialized.\n"
          ]
        }
      ],
      "source": [
        "# reload model\n",
        "ddqn.build_with_env(env)\n",
        "ddqn.load_model(\"d3rlpy_logs/ddqn.pt\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 20,
      "metadata": {},
      "outputs": [
        {
          "data": {
            "text/plain": [
              "Text(0, 0.5, 'episode reward')"
            ]
          },
          "execution_count": 20,
          "metadata": {},
          "output_type": "execute_result"
        },
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "# we see that training RL policy can be risky online\n",
        "log_path = sorted(glob(\"d3rlpy_logs/DoubleDQN_online_*/evaluation.csv\"))[-1]\n",
        "df = pd.read_csv(\n",
        "    log_path, \n",
        "    usecols=[1,2],\n",
        "    names=[\"train_episode\", \"episodic_rewards\"]\n",
        ")\n",
        "plt.plot(df[\"train_episode\"], df[\"episodic_rewards\"])\n",
        "plt.title(\"Learning curve of DDQN\")\n",
        "plt.xlabel(\"episode\")\n",
        "plt.ylabel(\"episode reward\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 23,
      "metadata": {},
      "outputs": [],
      "source": [
        "random = DiscreteRandomPolicyConfig().create(device=device)\n",
        "random.build_with_env(env)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 24,
      "metadata": {},
      "outputs": [
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "8cfab0e71ab34f68af1a98db06e35b5f",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "[calculate on-policy policy value]:   0%|          | 0/100 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "e0f25ddb71b54f88a7d55f6d6e7f72cc",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "[calculate on-policy policy value]:   0%|          | 0/100 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 400x400 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "# compare the policy performance online\n",
        "visualize_on_policy_policy_value(\n",
        "    env=env,\n",
        "    policies=[ddqn, random],\n",
        "    policy_names=[\"ddqn\", \"random\"],\n",
        "    n_trajectories=100,\n",
        "    random_state=random_state,\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 25,
      "metadata": {},
      "outputs": [
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "8708b0bdfbb74cde938d5727a8e6be50",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "[calculate on-policy policy value]:   0%|          | 0/100 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "35996a8d7a9d4156a0b096c19ec4455b",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "[calculate on-policy policy value]:   0%|          | 0/100 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "improvement: 1.558461332832902\n"
          ]
        }
      ],
      "source": [
        "# on-policy policy value by ddqn policy\n",
        "# largely outperforms the random policy (about 20%)\n",
        "ddqn_performance = calc_on_policy_policy_value(env, ddqn, n_trajectories=100, random_state=random_state)\n",
        "random_performance = calc_on_policy_policy_value(env, random, n_trajectories=100, random_state=random_state)\n",
        "print(\"improvement:\", ddqn_performance - random_performance)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "Note that, we provide the following functions for online evaluation of the policy performance under `scope_rl.ope.online`.\n",
        "\n",
        "(Statistics)\n",
        "\n",
        "* `calc_on_policy_policy_value`\n",
        "* `calc_on_policy_policy_value_interval`\n",
        "* `calc_on_policy_variance`\n",
        "* `calc_on_policy_conditional_value_at_risk`\n",
        "* `calc_on_policy_policy_interquartile_range`\n",
        "* `calc_on_policy_cumulative_distribution_function`\n",
        "\n",
        "(Visualization)\n",
        "* `visualize_on_policy_policy_value`\n",
        "* `visualize_on_policy_policy_value_with_variance`\n",
        "* `visualize_on_policy_cumulative_distribution_function`\n",
        "* `visualize_on_policy_conditional_value_at_risk`\n",
        "* `visualize_on_policy_interquartile_range`"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 1. Synthetic Dataset Generation\n",
        "In the previous section, we see how we can interact with the environment *online* using synthetic simulator.  \n",
        "Though simulation is a good starting point for training an agent, eventually we want to deploy our agent (or policy) in the real world environment.  \n",
        "However, online interaction may have potential risk of failure in the real-world.\n",
        "Therefore, we introduce an alternative learning protocol, Offline RL, which aims at training a new policy using only offline logged dataset collected by a behavior policy $\\pi_0$:\n",
        "\n",
        "$$\\mathcal{D}_0 := \\{ \\{ s_t, a_t, s_{t+1}, r_t \\}_{t=1}^T \\}_{i=1}^n \\sim \\prod_{i=1}^n d_{s_0}(s_0) \\prod_{t=1}^T \\pi_0(a_t | s_t) p(s_{t+1} | s_t, a_t) p(r_t | s_t, a_t),$$\n",
        "\n",
        "where $s_t \\in \\mathcal{S}$ is the state observed at timestep $t$, $a_t$ is the action chosen by the behavior policy $\\pi_0$, and $r_t$ is the reward. \n",
        "\n",
        "We prepare an easy-to-use synthetic dataset generator: `SyntheticDataset` class in the dataset module.  \n",
        "By regarding the simulation environment as a hypothetical real-environment, we show how to conduct Offline RL using logged dataset in real-world application.  \n",
        "\n",
        "\n",
        "`SyntheticDataset` has the following arguments:\n",
        "- `env`: Recommender environment for RL defined in the previous section.\n",
        "- `max_episode_steps`: Maximum number of timesteps in an episode."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 26,
      "metadata": {},
      "outputs": [],
      "source": [
        "# import necessary module from SCOPE-RL\n",
        "from scope_rl.policy import EpsilonGreedyHead"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 27,
      "metadata": {},
      "outputs": [],
      "source": [
        "behavior_policy = EpsilonGreedyHead(\n",
        "    ddqn, \n",
        "    n_actions=env.action_space.n,\n",
        "    epsilon=0.3,\n",
        "    name=\"ddqn_epsilon_0.3\",\n",
        "    random_state=random_state,\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 28,
      "metadata": {},
      "outputs": [],
      "source": [
        "# initialize dataset class\n",
        "dataset = SyntheticDataset(\n",
        "    env=env,\n",
        "    max_episode_steps=env.step_per_episode,\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "# collect logged data by a behavior policy\n",
        "# skip if there is a preserved logged dataset\n",
        "train_logged_dataset = dataset.obtain_episodes(\n",
        "    behavior_policies=behavior_policy,\n",
        "    n_trajectories=10000, \n",
        "    obtain_info=False,\n",
        "    random_state=random_state,\n",
        ")\n",
        "test_logged_dataset = dataset.obtain_episodes(\n",
        "    behavior_policies=behavior_policy,\n",
        "    n_trajectories=10000, \n",
        "    obtain_info=False,\n",
        "    random_state=random_state + 1,\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 30,
      "metadata": {},
      "outputs": [
        {
          "data": {
            "text/plain": [
              "{'size': 100000,\n",
              " 'n_trajectories': 10000,\n",
              " 'step_per_trajectory': 10,\n",
              " 'action_type': 'discrete',\n",
              " 'n_actions': 5,\n",
              " 'action_dim': None,\n",
              " 'action_meaning': None,\n",
              " 'action_keys': None,\n",
              " 'state_dim': 5,\n",
              " 'state_keys': None,\n",
              " 'state': array([[ 0.04479014,  0.11492637, -0.05131316, -0.0280654 ,  0.0347718 ],\n",
              "        [ 0.39224767,  0.80005173, -0.28628983, -0.17037224,  0.30833493],\n",
              "        [ 0.42240998,  0.78247161, -0.15893681, -0.18787797,  0.38567985],\n",
              "        ...,\n",
              "        [-0.4762458 , -0.61565697,  0.34349606, -0.11468651, -0.51284881],\n",
              "        [-0.42601396, -0.52915466,  0.40759172, -0.03914639, -0.60896962],\n",
              "        [-0.47442487, -0.57688529,  0.31635916, -0.01311537, -0.58469588]]),\n",
              " 'action': array([0, 1, 2, ..., 2, 0, 2]),\n",
              " 'reward': array([-0.07510555, -0.34569477,  0.11200114, ..., -0.20323486,\n",
              "         0.54777489, -0.28543168]),\n",
              " 'done': array([0., 0., 0., ..., 0., 0., 1.]),\n",
              " 'terminal': array([0., 0., 0., ..., 0., 0., 1.]),\n",
              " 'info': {},\n",
              " 'pscore': array([0.76, 0.06, 0.76, ..., 0.06, 0.76, 0.06]),\n",
              " 'behavior_policy': 'ddqn_epsilon_0.3',\n",
              " 'dataset_id': 0}"
            ]
          },
          "execution_count": 30,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "train_logged_dataset"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 29,
      "metadata": {},
      "outputs": [],
      "source": [
        "with open(\"logs/train_dataset_discrete_ddqn.pkl\", \"wb\") as f:\n",
        "    pickle.dump(train_logged_dataset, f)\n",
        "with open(\"logs/test_dataset_discrete_ddqn.pkl\", \"wb\") as f:\n",
        "    pickle.dump(test_logged_dataset, f)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 31,
      "metadata": {},
      "outputs": [],
      "source": [
        "with open(\"logs/train_dataset_discrete_ddqn.pkl\", \"rb\") as f:\n",
        "    train_logged_dataset = pickle.load(f)\n",
        "with open(\"logs/test_dataset_discrete_ddqn.pkl\", \"rb\") as f:\n",
        "    test_logged_dataset = pickle.load(f)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 32,
      "metadata": {},
      "outputs": [
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "5e5726f843f9481db9d22640874ff6cb",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "[calculate on-policy policy value]:   0%|          | 0/100 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "79138ded392d48d9bad3299bdfb1077a",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "[calculate on-policy policy value]:   0%|          | 0/100 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "3a4e8ebc5b9c4ff6b9f9cc68480c212d",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "[calculate on-policy policy value]:   0%|          | 0/100 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAi4AAAFjCAYAAAAaZIDwAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAAA7ZklEQVR4nO3de1yUdd7/8fcAg4KcBRUVRcEDeD7jMfOYpu5qabqdts1O2mbd+7PWrTzupm6PXXdLfdxb3ZWtJ8yy1dQ0UxEPiJrH8ixmIgYEioA44MzvD2/mlhhkZhjAkdfzH+f6Xt/rms8wg/Pme32v6zJYLBaLAAAA3IBHdRcAAABgL4ILAABwGwQXAADgNgguAADAbRBcAACA2yC4AAAAt0FwAQAAboPgAgAA3AbBBQAAuA2v6i7gXpOdna2ioqLqLgMAALfh5eWl4OBg+/pWci01TlFRkQoLC6u7DAAA7kkcKgIAAG6D4AIAANwGwQUAALgNggsAAHAbBBcAAOA2CC4AAMBtEFwAAIDbILgAAAC3QXABAABugyvnotrMmTNHWVlZkqSQkBC9+eab1VwRAOBuR3BBtcnKylJmZmZ1lwEAcCMcKgIAAG6D4AIAANwGwQUAALgNggsAAHAbTk/OLSgoUGpqqq5duyZJCggIUHh4uHx8fFxWHAAAwO0cCi7p6enavn279u/frx9//FFms7nEeg8PDzVu3FjdunXTfffdp/r167u0WAAAULPZFVwuXryo+Ph4JScnq06dOoqNjVVcXJzq16+vOnXqSJJyc3OVnp6uc+fOadOmTfrss8/UvXt3PfLII2rcuHGlvggAAFAz2BVcpk6dqk6dOmnatGlq166dPD0979j/5s2bOnr0qDZv3qypU6dqxYoVLikWAADUbHYFl7ffftuhURNPT0917NhRHTt2VGpqqtPFAQAA3M6us4oqcqinUaNGTm8LAABwO06HBgAAbsPus4o+/PBDh3ZsMBj01FNPOVwQAABAWewOLps2bXJ45wQXAADgSnYHl/j4+MqsAwAAoFzMcQEAAG7D7uBiNpv1xRdfKCEh4Y79EhIS9MUXX1S0LgAAgFLsDi47duzQypUrFRERccd+ERERWrlypRITEytcHAAAwO3sDi6JiYnq3Lmzmjdvfsd+zZs3V9euXcsdmQEAAHCU3cElJSVFbdu2tatvbGysUlJSnC4KAADAFruDy40bN+Tj42NXXx8fHxUUFDhdFAAAgC12B5eAgAClpaXZ1TctLU0BAQFOFwUAAGCL3cElJiZGiYmJunHjxh37FRQUKDExUbGxsRUuDgAA4HZ2B5dRo0bpypUrmjt3rrKysmz2ycrK0vz583XlyhWNHDnSZUUCAABIDlw5NzIyUhMnTtQHH3ygF198UTExMWrSpIl8fHx0/fp1XbhwQcePH5fFYtHTTz+tyMjISiwbAADURHYHF0kaOHCgIiIitHr1ah07dkzHjh2zrvPw8FDbtm01duxYtWzZ0uWFAgAAOBRcJKlly5b605/+JJPJpMuXLys/P1++vr5q0KCBvL29K6NGAAAASU4El2Le3t5q0qSJK2sBAAC4I26yCAAA3AbBBQAAuA2CCwAAcBtOz3G5mxQUFGjt2rU6ffq0zpw5o7y8PE2aNEn9+/cvd9vt27dr8eLFNte99957CgoKcm2xAADAafdEcMnJydHq1asVGhqqyMhIfffddw7vY9y4capXr16JNl9fX1eVCAAAXMAlweX7779XYmKisrKyFBQUpF69eqlDhw6u2LVdgoODraMjZ8+e1bRp0xzeR6dOnRQVFVUJ1QEAAFep8ByXr7/+WnPnzlVRUZGaNm2q3NxczZs3T+vWrXNFfXYxGo0uOaRz/fp1mc3mihcEAAAqhd0jLtevX5ePj0+p9g0bNmjKlCnq2rWrtW358uVav369W92vaNasWSooKJCXl5c6dOigJ554QuHh4dVdFgAAuI3dweWll17ShAkTNGDAgHL7GgwGGQyGChVWVby9vdW/f3+1adNGPj4+OnfunNavX6833nhD8+fPV2hoqM3tCgsLVVhYaF02GAzWYOcur/1uw88NAFAeu4PLxIkTtXTpUm3atElPPfWUWrduLUkaNmyY3nnnHcXFxSkkJESpqanat2+fJkyYUGlFu1KvXr3Uq1cv63L37t3VsWNHzZgxQ59//rmeffZZm9utWbNGq1evti43a9ZM8+fPV1hYWKXXfK/w9PQs8ZgRLgBAeewOLj169FDnzp21du1avfXWW+rcubMef/xxDRkyRA0aNNDu3buVkpKiwMBAvfrqq+rcuXNl1l2pWrdurejoaB09erTMPqNHj9aIESOsy8WjBRkZGSoqKqr0Gu8FN2/eLPE4LS2tGqsBAFQXLy8vu//wd+isIqPRqIceekj333+/li5dqpdfflkjR47Ur3/9a7Vv396pYu9WdevW1aVLl8pcbzQaZTQaba6zWCyVVdY9jZ8bAKA8Tp1VFBISopdeeklvvPGGDh48qClTpmjnzp2urq1apaenKyAgoLrLAAAAt3FoxCUzM1OHDh2SyWRSdHS0WrVqpbfeekvbtm3TJ598Yp3/0rx588qqt0Kys7OVn5+v+vXry8vr1kvPyckpFVC+/fZbnTt3TsOGDauOMgEAQBnsDi4HDhzQggULFBwcLF9fX33yyScaPny4nnjiCQ0YMEA9e/bUp59+qjfffFO9e/fWo48+qsDAwMqsvYSvvvpKeXl5ys7OliTt379fP//8s6RbE4h9fX21fPlyJSQkaOHChdar5L7xxhuKjIxUVFSUfH19lZKSom3btqlu3boaPXp0ldUPAADKZ3dwWbZsmeLi4vTiiy9KkhITE7Vw4UKNGDFCISEh8vHx0RNPPKFBgwZpyZIleumll7RkyZJKK/yX1q1bp4yMDOtycnKykpOTJUl9+/Yt8/L9vXr10rfffqsjR47oxo0bCg4O1sCBA/Xwww9znyIAAO4ydgeXn3/+ucShk1atWkmSsrKyFBISYm1v2LChpk2bpoMHD7qwzPItWrSo3D6TJ0/W5MmTS7SNHz9e48ePr6yyAACAC9kdXFq3bq2NGzcqIiJCderU0eeffy4/Pz9FRETY7N+pUyeXFQkAACA5EFyee+45LVq0SDNmzJAk1a9fX6+88opq1apVacUBAADczu7gEhISojfffFMmk0mFhYWqU6dOZdYFAABQikOnQ0u37u3j7e1dGbUAAADckcPB5ZfOnDmjI0eOKDc3VwEBAerYsaMiIyNdUBoAAEBJdgeXuXPnatSoUWrTpo2kW/eWeffdd7Vnz54S/VasWKHBgwdr4sSJrq0UAADUeHYHl0OHDqlv377W5dWrV2vPnj0aOHCgRowYodDQUF2+fFmff/65vv76azVr1kwDBw6slKIBAEDN5NS9iiTpm2++UdeuXfXss8+qYcOG8vb2VpMmTfTyyy+rZcuW2rJliyvrBAAAcC64FBQU6OrVq4qLi7O5vkePHkpNTa1QYQAAAL/kVHDx9vaWl5dXmddw8fb2lsViqVBhAAAAv+TQWUXLly/XF198YV2+cOGCunfvXqrf5cuXuc8PgLvGnDlzlJWVJen/rkkFwD3ZHVxiYmJkMBisy/7+/srMzCzVr6ioSLt27VJMTIxrKgSACsrKyrL5/xUA92N3cJk5c6Zd/QoLCzVlyhSFhoY6WxMAAIBNFb4A3S/5+PgoNjbW1butEdKm1qxr39zMunHb48wa9/rD3/6guksAALfjVHC5ceOGfvzxR2VlZclkMql27doKDw9Xo0aNXF0fAACAlUPBJS0tTStWrNCBAwdUVFRUan3dunX14IMPatiwYfLwcPoSMQAAADbZHVzOnz+vWbNmyWw2q127dqpVq5bOnDmjK1euaNSoUTKbzTpy5Ig++eQTHT16VFOnTpWnp2dl1g4AAGoYu4PLsmXL5Ofnpzlz5lhPdS4qKtLChQt1+PBhvfXWW5owYYJ2796td999V+vXr9eoUaMqq24AAFAD2X0859SpUxoyZEiJ67N4eXlpzJgxOnv2rH788UdJUq9evXTfffdp+/btrq4VAADUcA5NRLE1b6W4LT8/39rWsmVLpaenV7A0AACAkuwOLi1bttTXX3+tvLw8a5vFYtHatWvl5eWliIgIa3tubq58fHxcWykAAKjx7J7jMmHCBE2fPl0vvfSS2rVrJ29vb50+fVqXLl3S6NGj5evra+17+PBhNWvWrFIKBgAANZfdwaV58+aaPXu24uPjdfDgQRUVFalhw4aaOHGiBg8eXKLvww8/rJCQEJcXCwAAajaHruPSvHlzTZs2rdx+3KcIAABUBq4SBwAA3AbBBQAAuA2CCwAAcBsEFwAA4DYILgAAwG0QXAAAgNsguAAAALfhdHC5du2aHnnkER07dqzUusTERD3yyCMVKgwAAOCXGHEBAABug+ACAADcBsEFAAC4DYfuVSRJq1evliTduHFDkpSQkKATJ05Y1z/88MMuKg0AAKAkh4PLtm3bJEkWi0WSdPjwYRmNRkmSwWAguAAAgErjcHBZtGiRJCknJ0fPPPOMXnrpJbVt29blhQEAAPyS03NcDAaDK+sAAAAoF5NzAQCA2yC4AAAAt+F0cPHx8dELL7ygxo0bl1rXqlUrvfDCCxUqDAAA4Jccnpxr3dDLS/3797e5rl69eqpXr56zu3ZYQUGB1q5dq9OnT+vMmTPKy8vTpEmTyqzvl/Ly8rR06VIlJyfLZDIpOjpajz/+uJo3b165hQMAAIfcE4eKcnJytHr1aqWmpioyMtKhbc1ms+bNm6edO3fqgQce0KOPPqqrV69q1qxZSktLq5yCAQCAU+wKLjt37rRet8URFotFO3fudHg7RwUHB+u9997T4sWL9dhjjzm0bVJSkk6ePKnJkydr7NixeuCBBzRz5kx5eHho1apVlVQxAABwhl2HipYsWaJVq1Zp4MCB6tmzZ7mHgS5fvqzdu3dr27ZtKigoUJ8+fVxSbFmMRqOCgoKc2jYpKUmBgYHq3r27tS0gIEA9e/ZUYmKiCgsLrRfYAwAA1cuu4PLuu+9qw4YN+vLLL7V8+XLVq1dPzZo1U7169VSnTh1ZLBbl5eUpPT1d586dU2Zmpvz9/TVs2DA9+OCDlf0aKuT8+fNq3ry5PDxKDj5FR0dry5YtSktLU5MmTUptV1hYqMLCQuuywWCQj4+P9TFQHj4n1YefPeC+7AoutWvX1pgxY/SrX/1KBw4c0L59+3Tq1Cnt3bu3RL8GDRooJiZG3bp1U5cuXeTl5fTc3yqTnZ2tmJiYUu3FIzhZWVk2g8uaNWus922SpGbNmmn+/PkKCwtzupZLTm8JdxQeHl7dJdQYnp6eJR7zswfcl0PJwtPTU927d7ceVjGbzcrNzZUk+fn5lRq1cAcmk8nmoSBvb2/reltGjx6tESNGWJeL/4LLyMhQUVFRJVSKew2Tv6vOzZs3SzzmZw/cXby8vOz+w79CQyIeHh4KCAioyC6qnbe3d4lDPsWKA0txgPklo9FY5twXZyYyo+bhc1J9+NkD7sv9hkhcLDg4WFeuXCnVXtwWEhJStQUBAIAy1fjgEhkZqXPnzslsNpdoP336tGrVqsWxcAAA7iI1KrhkZ2crNTW1xByUuLg4Xb16VcnJyda2nJwcJSUlqUuXLpwKDQDAXeTuP+3HTl999ZXy8vKUnZ0tSdq/f79+/vlnSdKwYcPk6+ur5cuXKyEhQQsXLrReiyYuLk4tWrTQ4sWLdfHiRfn7+2vz5s0ym80aN25ctb0eAABQ2j0TXNatW6eMjAzrcnJysnUUpW/fvvL19bW5nYeHh6ZNm6alS5dq48aNMplMioqK0qRJk9SwYcMqqR0AANinQsGlsLBQKSkpunr1qlq1alWtZxgtWrSo3D6TJ0/W5MmTS7X7+fnp+eef1/PPP18ZpQEAABdxOrhs2LBBn376qfLz8yVJb775ptq2baucnBy98sorevTRRzVgwACXFQoAAODU5Nxt27ZpyZIl6tixo1544YUS6wICAtSmTRvt3r3bJQUCAAAUcyq4fPnll+rataumTJmiLl26lFrfvHlz/fjjjxUuDgAA4HZOBZfLly+rU6dOZa738/Oz3goAAADAVZwKLr6+vsrJySlz/cWLF603KQQAAHAVp4JLp06d9M033ygvL6/Uuh9//FHffPONzUNIAAAAFeHUWUXjx4/X66+/rj/84Q/WgLJ9+3Zt3bpVe/fuVXBwsB5++GGXFgoAAOBUcAkJCdG8efO0YsUK69lDiYmJql27tnr37q1HH33U7e8aDQAA7j5OX8clMDDQetG2nJwcmc1mBQQEyMOjRt3+CAAAVCGXXPKf0RXAvfx2yZ7qLqFK3cy9YX2cmXujxr3+j5/sWd0lAC7jVHBZvXq1Xf2Y5wIAAFzJqeDy6aef2tWP4AIAAFzJqeASHx9fqs1sNiszM1NfffWVjh8/rj/96U8VLg4AAOB2LptJ6+HhoXr16umJJ55QeHi4PvzwQ1ftGgAAQJILg8vtYmJidPDgwcrYNQAAqMEqJbicPXtWBoOhMnYNAABqMKfmuCQkJNhsz8vL0/Hjx5WcnKwBAwZUqDAAAIBfciq4LF68uMx1/v7++tWvfsUZRQAAwOWcCi4LFy4s1WYwGFSnTh35+PhUuCgAAABbnAouYWFhrq4DAACgXNxYCAAAuA27RlweeeQRh3dsMBi0cuVKh7cDAAAoi13B5aGHHuL0ZgAAUO3sCi7jxo2r7DoAAADKxRwXAADgNpw6q6jYzz//rJSUFOXn58tisZRaf99991Vk9wAAACU4FVxMJpMWLVqkvXv32gwsxQguuJNAT4Mky22PAQC4M6eCy4oVK5ScnKzx48erZcuWmjVrliZPnqygoCBt2LBB2dnZmjx5sqtrxT1mUl3v6i4BAOBmnJrjkpSUpP79++vXv/61IiIiJEkhISFq3769/vjHP8rX11ebNm1yaaEAAABOBZecnBxFR0dLkry9b/3VXFBQYF3fo0cPJScnu6A8AACA/+NUcAkMDNS1a9ckSbVq1VKdOnV06dIl6/rr16/LZDK5pkIAAID/5dQcl+joaJ04ccK63KVLF61bt07BwcGyWCxav369WrZs6bIiAQAAJCeDy/Dhw7Vnzx4VFhbKaDTqkUce0alTp6x3ja5fv76eeuoplxYKAABgd3D529/+pn79+qlTp05q3bq1WrdubV0XGhqqBQsW6MKFC/Lw8FCjRo3k6elZKQUDAICay+7g8u233yo5OVm+vr7q2bOn+vTpo9jYWOt6Dw8PRUZGVkaNAAAAkhwILv/zP/+jpKQk7dq1S9u2bdM333yjkJAQ9enTR3369FHTpk0rs04AAAD7g0vt2rXVv39/9e/fXzk5Odq1a5d27typtWvXau3atWrcuLH69u2rPn36KDQ0tDJrBgAANZRTk3MDAgI0bNgwDRs2TOnp6dq5c6d27dqlFStWaOXKlWrVqpX69u2rQYMGubpeAABQg1XoJouSVK9ePY0ZM0ZjxozRDz/8oFWrVmn//v06ceIEwQUAALhUhYOLJGVnZ1sPHaWkpEiSoqKiXLFrAAAAK6eDS15enpKSkrRz506dOHFCZrNZ9evX10MPPaR+/fqpQYMGrqwTAADAseBiMpm0f/9+7dy5U4cPH1ZRUZECAgI0ePBg9evXz3r/oupQWFio+Ph4JSYmKjc3V02bNtX48ePVvn37O263atUqrV69ulS70WjUsmXLKqtcAADgBLuDy8KFC7Vv3z4VFBSoVq1a6tGjh/r27asOHTrIw8OpWx651KJFi7R3714NHz5c4eHh2r59u+bOnasZM2aUuFheWSZOnKjatWtbl++G1wQAAEqyO7js2rVL7du3V58+fdS9e3fVqlWrMutyyJkzZ7R792499thjGjVqlCSpX79++sMf/qClS5fqz3/+c7n7iIuLU0BAQGWXCgAAKsDu4PKvf/3rrv1iT0pKkoeHR4mzmLy9vTVgwACtWLFCmZmZdl1bJj8/Xz4+PjIYDJVZLgAAcJLdweVuDS2SlJKSovDwcPn6+pZoL55zc/78+XKDy4svvmg9DNatWzc98cQTCgoKKrN/YWGhCgsLrcsGg0E+Pj7Wx0B5+JygqvBZw73EJadDV7crV64oODi4VHtxW3Z2dpnb+vn56YEHHlDLli3l5eWlEydOaNOmTTp79qzmzp1bKgwVW7NmTYlJvc2aNdP8+fMVFhbm9Ou45PSWcEfh4eHVXQJqCD5ruJfcE8HFZDLJaDSWai9uM5lMZW47fPjwEstxcXGKjo7WO++8o82bN+vXv/61ze1Gjx6tESNGWJeL/6LJyMhQUVGRoy8BNVBaWlp1l4Aags8a7nZeXl52/+F/TwQXb2/vEodtihW3eXt7O7S/Pn366JNPPtHRo0fLDC5Go9FmWJIki8Xi0POhZuJzgqrCZw33knvinN+goCCbh4OK22wdRipP3bp1lZubW+HaAACA6zgVXN566y3t3LnzjodgqlJkZKTS0tKUn59fov306dPW9Y6wWCzKyMi4qyckAwBQEzkVXH766Se9++67euaZZ7Rw4UIdOXKkWoci4+LiZDabtWXLFmtbYWGhtm/frhYtWljPKMrMzFRqamqJbXNyckrtb/PmzcrJyVGHDh0qt3AAAOAQp+a4/POf/9SZM2eUmJioPXv2KDExUUFBQerTp4/69u3r8AhHRbVo0UJxcXFasWKFcnJy1KBBAyUkJCgjI0PPP/+8td/ChQv1/fffa9WqVda2SZMmqVevXmrSpImMRqNOnDih3bt3KzIyUoMHD67S1wEAAO7M6cm50dHRio6O1pNPPqkjR44oMTFRW7Zs0ZdffqnGjRurX79+6tOnj+rWrevKesv04osvKj4+Xjt27FBeXp6aNGmi1157TbGxsXfcrk+fPjp16pT27t0rk8mksLAwjRo1SmPGjLmrrg4MAAAkg8WFx3jy8vL03nvvKSkp6dbODQa1adNGDz74oDp37uyqp7mrZWRk2DzDyR5pUye6uBrczcLf/qDanvu3S/ZU23NXh5s7/y0VXLu1UNtfnn0er96CqtjHT/as7hKAOzIajVV7OvSJEye0Y8cO7d27V7m5uYqIiFC/fv3k5eWlbdu2af78+RozZoweeeQRVzwdAACooZwOLhcvXtSOHTu0a9cuZWZmKjAwUPfdd5/69etXYo7L8OHD9a9//UubNm0iuAAAgApxKrhMnTpVFy5ckNFoVNeuXTVx4kR16NBBHh62T1Jq06aNtm7dWqFCAQAAnAouderU0XPPPae4uLgy7+Vzu27dumnhwoXOPBUAAICVU8Fl5syZDvWvVatWhW4+CAAAIDl5Abpz585p06ZNZa7ftGmTzp8/72xNAAAANjkVXFauXKmjR4+Wuf7YsWNauXKl00UBAADY4vSIS+vWrctcHxMTo7NnzzpdFAAAgC1OBZfr16/L09OzzPUGg6HUDQ8BAAAqyqngEh4ersOHD5e5/tChQ6pfv77TRQEAANjiVHAZMGCADh48qCVLligvL8/anpeXp48//liHDh3SgAEDXFYkAACA5OTp0MOGDdP58+e1YcMGbdy4UcHBwZKk7OxsWSwW9e3bVw8++KBLCwUAAHAquBgMBk2aNEn9+vXT3r17lZ6eLunWheZ69OihNm3auLRIAKiQWnVsPwbgdip0k8W2bduqbdu2rqoFACqFZ7cx1V0CABdxao4LAABAdbBrxGXy5Mny8PDQggUL5OXlpcmTJ8tgMNxxG4PBoHfffdclRQIAAEh2BpfY2FgZDAbr3Z+LlwEAAKqS3SMud1oGAACoCsxxAQAAbsOuEZfvv//eqZ3HxsY6tR0AAIAtdgWXWbNmObXz+Ph4p7YDAACwxa7gMmPGjMquAwAAoFx2n1UEAABQ3Sp05VxJKigoUGZmpiQpNDRUtWvXrnBRAAAAtjgdXM6cOaNly5bpxIkTMpvNkiQPDw+1bt1ajz32mKKiolxWJAAAgORkcDl9+rRmzpwpLy8vDRgwQI0aNZIkpaamateuXZoxY4Zmzpyp6OholxYLAABqNqeCy8qVKxUSEqI5c+YoKCioxLqxY8fqzTff1IoVK/Tmm2+6okYAAABJTl6A7vTp0xo8eHCp0CJJQUFBGjRokE6fPl3R2gAAAEpwKrgYDAbdvHmzzPVms5l7GQEAAJdzKri0atVKmzZtUkZGRql1mZmZ2rx5s1q3bl3h4gAAAG7n1ByXCRMmaMaMGXr55ZfVvXt3hYeHS5IuXbqk/fv3y9PTUxMmTHBpoQAAAE4Fl2bNmukvf/mLVq5cqf3798tkMkmSvL291bFjR40fP16NGzd2aaEAAABOX8clIiJCU6dOldlsVk5OjiQpICBAHh7ccBoAAFQOh4LLli1btH79eqWnp8vPz089e/bUo48+avPsIgAAAFeze3gkOTlZ77//vn7++Wc1adJEFotFGzdu1Pvvv1+Z9QEAAFjZPeKyfv161a9fX7Nnz1ZQUJBu3rypd999V4mJifrtb38rX1/fyqwTAADA/hGXS5cuadCgQdbDQp6enho9erTMZrNSU1Mrqz4AAAAru4NLTk6OQkJCSrQVL9+4ccO1VQEAANjAKUAAAMBtOHRW0bp167Rr1y7rcvFl/1euXCl/f/8SfQ0Gg1599VUXlAgAAHCL3cElNDRUubm5ys3NLdWenZ2t7OzsEu3cqwgAALia3cFl0aJFlVlHhRUWFio+Pl6JiYnKzc1V06ZNNX78eLVv377cbbOysvTxxx/ryJEjslgsatOmjZ588knVr1+/CioHAAD2umfmuCxatEjr169Xnz599NRTT8nDw0Nz587ViRMn7rhdQUGBZs2apePHj2v06NEaO3asUlJSNHPmTF27dq2KqgcAAPa4J4LLmTNntHv3bk2YMEGPP/64Bg0apOnTpys0NFRLly6947abNm1SWlqaXnvtNf3qV7/SiBEj9MYbbyg7O1vr1q2rolcAAADscU8El6SkJHl4eGjQoEHWNm9vbw0YMECnTp1SZmbmHbeNiopSdHS0ta1Ro0Zq166d9uzZU6l1AwAAx9wTwSUlJUXh4eGlrt5bHEbOnz9vczuz2awLFy4oKiqq1LqoqCj99NNPun79usvrBQAAznH67tB3kytXrig4OLhUe3HbL894Kpabm6vCwkKbN4m8fVsfH59S6wsLC1VYWGhdNhgM1n6cUQV78DlBVeGzhnvJPRFcTCaTjEZjqfbiNpPJVOZ2t/dzZNs1a9Zo9erV1uVmzZpp/vz5CgsLc6z424Qv2+D0toAjNk97qLpLQA3x0X8fqO4SUIWeer5LpT+HU8Hl/fffV79+/dSqVStX1+MUb2/vEqMfxYrbvL29y9zu9n6ObDt69GiNGDHCulz8F01GRoaKioocqB4AgHtDWlqaU9t5eXnZ/Ye/U8Fl586d2rJli+rVq6e+ffuqb9++Cg8Pd2ZXLhEUFKSsrKxS7cWHiGwdRpIkPz8/GY1GXblyxeFtjUajzZEaSbJYLPaUDQDAPaUqvv+cCi4ffPCB9u/fr8TERH3xxRf67LPPFBUVpX79+qlXr14KCAhwdZ13FBkZqe+++075+fklJuiePn3aut4WDw8PNWnSRGfPni217syZM6pfv77N+S0AANiyaeti5V+/Kkny9QnU0AGTqrmie49TZxUZjUb17NlTr776qt577z1NnDhRRqNRH330kZ5//nnNmzdPu3fvLnN+iKvFxcXJbDZry5Yt1rbCwkJt375dLVq0UGhoqCQpMzNTqampJbbt0aOHzp49WyK8XLp0SceOHVNcXFyV1A8AuDfkX7+qvPxs5eVnWwMMXKvCk3P9/Pw0ePBgDR48WJmZmfr3v/+tpKQkHTx4UD4+PoqLi9OwYcPUtGlTV9RrU4sWLRQXF6cVK1YoJydHDRo0UEJCgjIyMvT8889b+y1cuFDff/+9Vq1aZW0bOnSovvnmG82bN08jR46Up6envvzySwUGBmrkyJGVVjMAAHCcS84qyszM1M6dO5WYmKiLFy/Kz89PvXr1kpeXlxITE7V9+3b97ne/05AhQ1zxdDa9+OKLio+P144dO5SXl6cmTZrotddeU2xs7B238/Hx0cyZM7VkyRJ99tlnslgsio2N1W9/+9sqP+QFAADuzGBxciZNXl6e9uzZo8TERJ08eVKenp7q1KmT+vXrp86dO8vL61YmKiws1D//+U+dOnVK7733nkuLvxtlZGTYPEsJAGqiTWudO8vEXa1ZP1d5+bdO7qjjG6zRD06r5oqq1tBRzp2oYzQaK/esorfffluHDh1SUVGRoqOj9bvf/U69evWSn5+fzWLi4uK0b98+Z54KAADAyqngcv78eY0cOVL33XefXadBt2/fXjNmzHDmqQAAAKycCi6LFi1yqH9AQEC5c00AAADK49Tp0Onp6dq/f3+Z6/fv36/09HSniwIAALDFqeDyySefaOPGjWWu37Rpk5YvX+50UQAAALY4FVxOnz6t9u3bl7m+Xbt2On78uNNFAQAA2OJUcMnNzb3jpfBr166t3Nxcp4sCAACwxangEhoaqhMnTpS5/vjx4woJCXG6KAAAAFucCi69e/fWrl27tGHDBpnNZmu72WzWhg0btHv3bvXp08dlRQIAAEhOng49evRonTx5UkuWLNGaNWvUsGFDSbduTpiTk6PY2FiNGTPGpYUCAAA4FVyMRqNef/11JSQkaO/evfrpp58kSVFRUYqLi1O/fv3k4eHUYA4AAECZnL7JooeHh+6//37df//9rqwHAACgTAyLAAAAt2HXiMusWbNkMBj0+uuvy9PTU7NmzSp3G4PBoOnTp1e4QAAAgGJ2jbhYLBZZLJYSy/ZsAwAA4Ep2jbjMnDnzjssAAABVgTkuAADAbdg14pKZmenUzkNDQ53aDgAAwBa7gsvkyZOd2nl8fLxT2wEAANhiV3B54YUXKrsOAACActkVXPr371/JZQAAAJTPJZNzTSaTTCaTK3YFAABQJqcv+Z+ZmalVq1bp4MGDysnJkSQFBASoU6dOGjt2rMLCwlxWJAAAgORkcElNTdX06dOVl5en9u3bq1GjRpJu3R16x44dOnDggObMmWO9azQAAIArOBVcli1bJoPBoL/+9a9q0qRJiXUXLlzQnDlztGzZMk2dOtUlRQIAAEhOznE5fvy4hg0bViq0SFKTJk00dOhQff/99xUuDgAA4HZOBZeioiJ5e3uXub5WrVoqKipyuigAAABbnAouzZo109atW5Wfn19qXX5+vrZu3armzZtXuDgAAIDbOTXHZdy4cfrLX/6il19+Wf3797dOwr106ZISEhJ07do1Pf300y4tFAAAwKng0rZtW02bNk1Lly7Vf/7znxLrIiMj9eKLL6pt27YuKRAAAKCY09dxad++vf7617/qypUrysjIkCSFhYUpKCjIVbUBAACU4FBwOXTokDZs2KD09HT5+/urZ8+eGj58OGEFAABUCbuDy/fff6+5c+dKkvz9/fXTTz/p1KlTysrK0mOPPVZpBQIAABSz+6yiNWvWKCgoSG+//bY++OADvf/++2rbtq02bdrEfYoAAECVsHvE5cKFCxoyZIj1onN+fn6aMGGCXn/9df3444+KioqqtCIBAO5p6Kjw6i6hSn21zVN5/3ulEB9fzxr3+quC3SMuV65cUb169Uq0FS9fv37dtVUBAADY4NAF6AwGwx2XAQAAKpNDZxUlJCTo1KlT1uXCwkJJ0ldffaXk5OQSfQ0Gg5566ikXlAgAAHCLQ8HlyJEjOnLkSKn2ffv22exPcAEAAK5kd3CJj4+vzDoAAADK5dRNFgEAAKqDS4JLfn6+Fi9erNTUVFfsDgAAwCan71V0O5PJpISEBPXr10+NGjVyxS4dkpeXp6VLlyo5OVkmk0nR0dF6/PHH1bx583K3XbRokRISEkq1N2zYUP/4xz8qoVoAAOAslwSX6mQ2mzVv3jydP39eo0aNkr+/vzZv3qxZs2Zp3rx5Cg8v/+I/RqNRzz33XIk2X1/fyioZAAA4ye2DS1JSkk6ePKn/+q//UlxcnCSpV69emjJlilatWqUpU6aUuw8PDw/169evsksFAAAV5JI5Ll5eXoqNjVWdOnVcsTuHJCUlKTAwUN27d7e2BQQEqGfPntq/f7/1WjPlMZvNys/Pr6wyAQCAC7hkxMXPz08zZsxwxa4cdv78eTVv3lweHiUzWHR0tLZs2aK0tDTr/ZXKYjKZ9OSTT+rGjRuqU6eOevfurccee0y1a9cuc5vCwsISochgMMjHx8f6GAAAvg9cr8LBpaCgQLm5uTbXhYaGVnT35crOzlZMTEyp9qCgIElSVlbWHYNLcHCwRo0apWbNmslisejQoUPavHmzfvjhB82cOVOenp42t1uzZo1Wr15tXW7WrJnmz5+vsLCwir0gAIDbuv07w9PT0655lnCMU8HFZDJp9erV2rp1q65du1ZmP0cvWmc2m1VUVGRXX6PRKIPBIJPJJKPRWGq9t7e3tdY7+c1vflNiuXfv3goPD9fKlSuVlJSk3r1729xu9OjRGjFihHW5OFVnZGTY/RoAAPeWmzdvlniclpZWjdW4Dy8vL7v/8HcquHzwwQdKSEhQt27dFBMT47K5LcePH9esWbPs6rtgwQI1atRI3t7eNuexFAeW4gDjiBEjRig+Pl5Hjx4tM7gYjUabgUmSLBaLw88JALj38H3gek4Fl+TkZA0cOFDPPvusS4tp1KiRJk2aZFff4OBg679Xrlwptb64LSQkxOE6vL295e/vX+YhMAAAUD2cCi4Gg0HNmjVzdS0KCgpS//79HdomMjJSx48fl9lsLjFB9/Tp06pVq5ZTxxevX7+ua9euKSAgwOFtAQBA5XHqdOiuXbvq6NGjrq7FKXFxcbp69aqSk5OtbTk5OUpKSlKXLl1KHM65fPmyLl++bF02mUy6fv16qX1+9tlnslgs6tixY6XWDgAAHOPUiMtDDz2kBQsW6F//+pcGDx6s0NDQUqcjS7dOk65scXFxatGihRYvXqyLFy9ar5xrNps1bty4En3nzJkj6dZl/qVbh5Nee+019e7dWw0bNpQkHT58WAcPHlTHjh3VtWvXSq8fAADYz6ngUnw12vPnz2vr1q1l9nP0rCJneHh4aNq0aVq6dKk2btwok8mkqKgoTZo0yRpGylKnTh117txZR44cUUJCgsxmsxo0aKAJEyZo5MiRNsMYAACoPgaLE1OeV61aZddFdcaOHetUUe4sIyPD7qv1AgDuLa+88ooyMzMl3bqW2YIFC6q5IvdgNBor93ToXx6CAQAAqAoVOhZSUFCg7OxsFRQUuKoeAACAMjk84pKenq61a9fq22+/1c8//2xtDwkJUZcuXTRq1CjVq1fPpUUCAABIDgaXffv2aeHChSooKFBYWJi6dOkiHx8fXb9+XRcuXNDXX3+tHTt26Pe//726detWWTUDAIAayu7gcvHiRf3jH/9QvXr19Oyzz9q8seHx48f1/vvv6x//+Ifmz5+vxo0bu7RYAABQs9k9x+Xzzz+Xv7+/5syZYzO0SFJMTIxmz54tf39/rVmzxmVFAgAASA4El++++04DBgwo96Jyfn5+uv/++3Xs2LEKFwcAAHA7u4NLbm6u3edY16tXjxsUAgAAl7M7uPj7+ys9Pd2uvunp6fL393e6KAAAAFvsDi6xsbHaunVruSMpubm52rp1q2JjYytcHAAAwO3sDi5jxoxRbm6uZsyYoZMnT9rsc/LkSc2YMUO5ubkaPXq0y4oEAACQHDgdunHjxnrppZe0cOFCTZ8+XfXq1VPTpk1Vu3ZtFRQU6IcfflB6erq8vb31+9//XhEREZVZNwAAqIEcugBdjx49FBkZqf/85z/69ttvtW/fPuu6oKAgDRw4UKNGjVKDBg1cXigAAIDDl/yvX7++nn32WUlSfn6+CgoKVLt2bfn6+rq8OAAAgNs5dXfoYr6+vgQWAABQZSoUXAAAwP8JCQmx+RiuY7BYLJbqLuJekpGRocLCwuouAwAAt2E0Gu2+yK3dp0MDAABUN4ILAABwGwQXAADgNgguAADAbRBcAACA2yC4AAAAt0FwAQAAboPgAgAA3AbBBQAAuA2CCwAAcBvcq8jFvLz4kQIA4AhHvju5VxEAAHAbHCpCtbp+/bpee+01Xb9+vbpLwT2OzxqqCp+1ykVwQbWyWCxKSUkRA3+obHzWUFX4rFUuggsAAHAbBBcAAOA2CC6oVkajUQ8//LCMRmN1l4J7HJ81VBU+a5WLs4oAAIDbYMQFAAC4DYILAABwGwQXAADgNgguAADAbRBc4FKrVq3SuHHjyu03btw4rVq1qgoqgivZ+/5K1f8enzlzRhMmTFBGRka11VBZDh06pMcff1w5OTnVXQrKUd2/B/ci7ggI4J60YsUK9e7dW2FhYVX6vFlZWfr444915MgRWSwWtWnTRk8++aTq169f7raff/65Dhw4oMuXL6ugoEB169ZV586dNWbMGAUEBFj7dezYUQ0aNNCaNWv05JNPVubLAe46jLgAuOecP39eR48e1ZAhQ6r0eQsKCjRr1iwdP35co0eP1tixY5WSkqKZM2fq2rVr5W5/7tw5NW3aVGPGjNHTTz+trl27atu2bXrjjTdUUFBQou/gwYO1ZcsW7oeDGocRFwD3nG3btik0NFQtWrSo0ufdtGmT0tLS9NZbbyk6OlqS1KlTJ/3hD3/QunXr9Jvf/OaO2/+///f/SrW1bNlSf//733XgwAH17t3b2t6jRw99+OGH2rNnjwYMGODaF3IPKygoUO3atau7DFQAwQVOO3HihJYsWaILFy4oJCREo0aNKtWnsLBQy5YtU2JiogoLC9WmTRtNnDjR7v1lZ2dr9erVJY4Rjxs3TkOHDlW7du0UHx+vtLQ0NWjQQE888YQ6duxYWS+3xrHn/ZXuzvd43759atu2rQwGQ6l1Bw8e1Jo1a5SSkiKDwaCYmBg99thjioiIsO8HcwdJSUmKioqyhhZJatSokdq1a6c9e/aUG1xsqVevniQpLy+vRHtgYKCaNm2q/fv3E1zKsGrVKq1evVp///vf9dlnn+nQoUMKCwvT5MmT9eWXX+r48ePKzs6Wr6+vOnXqpMcff1z+/v6ltn/nnXf02Wefad++fbJYLOrRo4eefvpp1apVy9rXkd+DlJQUrVixQidPnpTZbFaLFi00fvx4tWzZ0tpn+/btWrx4sWbPnq3du3dr165dunnzpnr16qXf/e53unHjhj766CMdOHBAkjRw4EA9+uijNj/z9xqCC5xy4cIF/fnPf1ZAQIDGjh2rmzdvatWqVQoKCirR77//+7+VmJioPn36qGXLljp27JjmzZvn9P6KnThxQsnJyRoyZIh8fHy0ceNG/e1vf9PixYtL/McD5zjyftxt73FWVpYyMzPVrFmzUut27NihRYsWqUOHDnr00Ud148YNbd68WdOnT9f8+fOtIaGwsNDuQzDFc0/MZrMuXLig+++/v1SfqKgoHT58WNevX5ePj88d92exWHTt2jWZzWalpaVp+fLl8vDwUJs2bUr1bdasmfbt22dXnTXZ3//+dzVo0EATJkyQxWLRkSNHlJ6erv79+ysoKEgXL17Uli1bdPHiRf3lL38p9eW/YMEChYWF6Te/+Y3OnTunrVu3KiAgQI899pi1j72/Bz/++KOmT58uX19fjRo1Sp6entqyZYtmzZqlmTNnlhol/PDDDxUUFKRx48bp9OnT2rJli3x9fXXq1CmFhoZqwoQJ+vbbb7V27VpFRETovvvuq5wf4l2E4AKnxMfHy2KxaPbs2QoNDZV0a+j69qHu8+fPKzExUUOGDLH+5fHAAw/onXfe0Q8//ODw/m6Xmppq/c9Iktq0aaOpU6dq165deuCBB1z+emsae9+Pu/E9Tk1NlfR/IxXFCgoK9NFHH2nAgAF67rnnrO333XefXn75Za1Zs8bavmvXLi1evPjOP6T/VTxSlJubq8LCQptBLDg4WJKUnZ1dbnC5evWqnn32Wety3bp19dJLL6lRo0al+tavX1/Xrl3T1atXFRgYaFe9NVHTpk01ZcoU67LJZNLIkSNL9GnRooX++c9/6sSJE4qJiSmxLjIyUi+88IJ1OTc3V9u2bbMGF0d+D1auXKmbN29q9uzZ1gnbxZ/BpUuXatasWSX6BwYGatq0aTIYDBo6dKguX76sdevWadCgQXrmmWckSYMGDdLkyZO1bds2ggtgi9ls1uHDh9WtWzfrF5AkNW7cWB06dNDBgwclyfrv8OHDS2w/fPhw7dy50+H93a5du3bWLzTp1n9MPj4++umnn1zzImswR96Pu/E9Lp4EW6dOnRLtR44cUV5ennr37l3iNGIPDw+1aNFC3333nbWtQ4cOeuONN+74PL9kMpkkyeaN9YrbivvciZ+fn9544w0VFhYqJSVFycnJpSbmFit+jdeuXSO43MHgwYNLLHt7e1sfm0wmFRQUWEc6UlJSSgWXX27funVrJScnKz8/X76+vg79Hhw5ckTdunUrcZZZcHCwevfurW+++ca6z2IDBgwoMQIUHR2tU6dOlTg86OHhoebNm+vcuXP2/UDcHMEFDsvJyZHJZFJ4eHipdQ0bNrT+EmdkZMhgMJQ6DbRhw4ZO7e92t3/5FfPz8ys1DwCOc+T9cKf3OC0tTZI0e/Zsm+tvHwkJDg62jpLYq/jLsLCwsNS64rbbvzDL4uXlpfbt20uSunTponbt2unNN99UYGCgunTp4lBNuOWXo2+5ubn69NNPtXv3bl29erXEuvz8/FLb//Kz6OfnJ+nWvCNfX1+Hfg9u3LhRql26FeItFot+/vnnEsHll89dvK5u3bql2mvK/38EF7glDw/bZ/Jzs/N7h7PvcfH8l9zcXJvbvfjiizYP53h6elofm0wmm19gthTvy8/PT0ajUVeuXCnVJzs7W5IcDkOS1KpVKwUHBysxMbFUcCl+jbdf4wWl/TIwLliwQCdPntSoUaMUGRmp2rVry2w266233pLZbC61fXX+f1PWc9tqryn//xFc4LCAgAB5e3tb/4K93aVLl6yPw8LCZLFY9NNPP5X4C+P2Po7sD1XDkffjbnyPi+eCpKenl2gv/ms4MDDQOqJRlt27dzs8x8XDw0NNmjTR2bNnS/U5c+aM6tevX+78lrKYTCabk4XT09Pl7+9PcHFAbm6ujh49qnHjxunhhx+2ttv6bNrLkd+DWrVq2fzMp6amymAwlBpJQWkEFzjMw8NDHTp00L59+5SZmWkdyrx48aIOHz5s7depUyetWLFCGzZsKHFa4IYNG5zaH6qGI+/H3fgeh4SEqG7duqWO93fo0EE+Pj5as2aN2rRpIy+vkv/95eTkWAOAM3NcpFuTjZcvX66zZ88qKipK0q0vr2PHjpWaDJqamqpatWpZfxYFBQUyGAwlTrGVbp1inZeXp+bNm5d6vpSUlBKn0KJ8xSMVvxydWL9+vdP7dOT3oH379tq/f7/S09Oth7CuXLminTt3qnXr1iUOE8E2ggucMm7cOB06dEjTp0/XkCFDZDabtXHjRkVERFhn0UdGRqp3797avHmz8vPz1apVKx09etTm5Ep79oeqY+/7cbe+x926dVNycrIsFot1YqOvr6+eeeYZvfvuu3rttdfUu3dvBQQEKDMzU99++61atWqlp59+WpJzc1wkaejQofrmm280b948jRw5Up6envryyy8VGBhYKri88sorio2N1cyZMyVJly9f1pw5c9SzZ081atRIBoNB586dU2JiosLCwkpN/Lx69ap++OEHDR061ImfUM3l6+urmJgYrV27Vjdv3lRISIgOHz5caoTOEY78HowfP15Hjhyx/h4Unw5dVFRU4vRqlI1L/sMpTZs21euvv66AgACtWrVK27Zt07hx49StW7cS/V544QUNGzZMhw8f1tKlS3Xz5k398Y9/dHp/qBqOvB9343t8//33KysrSydPnizR3qdPH02fPl0hISFau3atPvroI+3atUuRkZE2r7/iKB8fH82cOVMxMTH67LPPFB8fr6ZNm2rWrFnlHs4JCQlRjx499N1332n58uX65JNPdPLkSQ0dOlRz584tde2avXv3ymg0qmfPnhWuu6aZMmWKOnTooE2bNmn58uXy9PTUn/70pwrt097fg4iICM2ePVsRERH64osvtHr1aoWGhmrGjBlVfqVnd2Ww1JTZPHBLxVeu5O6q967Keo9nz56t4OBg/f73v3fpfu8Wr776qmJjY/Xb3/62uksBqhQjLgDuSRMmTNDu3buVkZFR3aW43KFDh5SWlqbRo0dXdylAlWOOC4B7UosWLbRixYrqLqNSdOzYUf/+97+ruwygWjDiAgAA3AZzXAAAgNtgxAUAALgNggsAAHAbBBcAAOA2CC4AAMBtEFwAAIDbILgAAAC3QXABAABug+ACAADcxv8HLQCR0xzC334AAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 600x400 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "# compare the policy performance online\n",
        "# almost the same performance as greedy ddqn\n",
        "visualize_on_policy_policy_value(\n",
        "    env=env,\n",
        "    policies=[ddqn, behavior_policy, random],\n",
        "    policy_names=[\"ddqn\", \"ddqn (e=0.3)\", \"random\"],\n",
        "    n_trajectories=100,\n",
        "    random_state=random_state,\n",
        ")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "For more about the data collection and visualization, please refer to [examples/quickstart/rec/rec_synthetic_data_collection.ipynb](https://github.com/hakuhodo-technologies/scope-rl/blob/main/examples/quickstart/rec/rec_synthetic_data_collection.ipynb)."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 2. Offline Reinforcement Learning\n",
        "In this section, we learn a new policy in an offline manner, without interacting with the environment.  \n",
        "We use [d3rlpy](https://github.com/takuseno/d3rlpy)'s implementation of algorithms."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 33,
      "metadata": {},
      "outputs": [],
      "source": [
        "# import modules from d3rlpy\n",
        "from d3rlpy.dataset import MDPDataset\n",
        "from d3rlpy.algos import DiscreteCQLConfig"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 34,
      "metadata": {},
      "outputs": [],
      "source": [
        "offlinerl_dataset = MDPDataset(\n",
        "    observations=train_logged_dataset[\"state\"],\n",
        "    actions=train_logged_dataset[\"action\"],\n",
        "    rewards=train_logged_dataset[\"reward\"],\n",
        "    terminals=train_logged_dataset[\"done\"],\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 35,
      "metadata": {},
      "outputs": [],
      "source": [
        "# Use Conservative Q-Learning policy\n",
        "cql = DiscreteCQLConfig(\n",
        "    encoder_factory=VectorEncoderFactory(hidden_units=[30, 30]),\n",
        "    q_func_factory=MeanQFunctionFactory(),\n",
        ").create(device=device)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "cql.fit(\n",
        "    offlinerl_dataset,\n",
        "    n_steps=10000,\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 37,
      "metadata": {},
      "outputs": [],
      "source": [
        "# save model\n",
        "cql.save_model(\"d3rlpy_logs/cql_discrete.pt\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 39,
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-07-22 18:02:19 [warning  ] Parameters will be reinitialized.\n"
          ]
        }
      ],
      "source": [
        "# reload model\n",
        "cql.build_with_env(env)\n",
        "cql.load_model(\"d3rlpy_logs/cql_discrete.pt\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 40,
      "metadata": {},
      "outputs": [
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "a607a1aba17641f990ae8cfbd1e3c625",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "[calculate on-policy policy value]:   0%|          | 0/100 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "11b098763c8e40c28e22892c8a65b052",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "[calculate on-policy policy value]:   0%|          | 0/100 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "56a1b9a5f5624f869af9ba1c38e6417b",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "[calculate on-policy policy value]:   0%|          | 0/100 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "6cb3aa5bbb74407c82641c8768bef6aa",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "[calculate on-policy policy value]:   0%|          | 0/100 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 800x400 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "# compare the policy performance online\n",
        "# not better than online policy, but achieves almost the same performance\n",
        "# meaning that we can learn a reasonable policy without interacting with the environment!\n",
        "visualize_on_policy_policy_value(\n",
        "    env=env,\n",
        "    policies=[ddqn, behavior_policy, cql, random],\n",
        "    policy_names=[\"ddqn\", \"ddqn (e=0.3)\", \"cql (offline)\", \"random\"],\n",
        "    n_trajectories=100,\n",
        "    random_state=random_state,\n",
        ")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 3. Off-Policy Evaluation and Evaluation of Offline RL methods\n",
        "In the previous section, we verified the performance of a learned policy via online interaction (aka A/B tests).  \n",
        "However, it could be risky because the policy may perform poorly in an online environment.  \n",
        "To tackle this issue, we aim at evaluating the performance of a policy offline, to gain some prior knowledge on how agent works in the environment.\n",
        "\n",
        "#### Estimation Target\n",
        "\n",
        "The goal to estimate the estimators' average policy performance, the *policy value*.\n",
        "\n",
        "$$ V(\\pi) := \\mathbb{E}\\left[\\sum_{t=1}^T \\gamma^{t-1} r_t \\mid \\pi \\right] $$"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 47,
      "metadata": {},
      "outputs": [],
      "source": [
        "# import ope modules from SCOPE-RL\n",
        "from scope_rl.ope import CreateOPEInput\n",
        "from scope_rl.ope import OffPolicyEvaluation as OPE\n",
        "from scope_rl.ope.discrete import DirectMethod as DM\n",
        "from scope_rl.ope.discrete import TrajectoryWiseImportanceSampling as TIS\n",
        "from scope_rl.ope.discrete import PerDecisionImportanceSampling as PDIS\n",
        "from scope_rl.ope.discrete import DoublyRobust as DR\n",
        "from scope_rl.ope.discrete import SelfNormalizedTIS as SNTIS\n",
        "from scope_rl.ope.discrete import SelfNormalizedPDIS as SNPDIS\n",
        "from scope_rl.ope.discrete import SelfNormalizedDR as SNDR"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 48,
      "metadata": {},
      "outputs": [],
      "source": [
        "# define evaluation policies\n",
        "cql_ = EpsilonGreedyHead(base_policy=cql, n_actions=env.action_space.n, name=\"cql\", epsilon=0.0, random_state=random_state)\n",
        "ddqn_ = EpsilonGreedyHead(base_policy=ddqn, n_actions=env.action_space.n, name=\"ddqn\", epsilon=0.0, random_state=random_state)\n",
        "random_ = EpsilonGreedyHead(base_policy=random, n_actions=env.action_space.n, name=\"random\", epsilon=1.0, random_state=random_state)\n",
        "evaluation_policies = [cql_, ddqn_, random_]"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 43,
      "metadata": {},
      "outputs": [],
      "source": [
        "# first, prepare OPE inputs\n",
        "prep = CreateOPEInput(\n",
        "    env=env,\n",
        "    model_args={\n",
        "        \"fqe\": {\n",
        "            \"encoder_factory\": VectorEncoderFactory(hidden_units=[30, 30]),\n",
        "            \"q_func_factory\": MeanQFunctionFactory(),\n",
        "            \"learning_rate\": 1e-4,\n",
        "        }\n",
        "    },\n",
        "    state_scaler=MinMaxObservationScaler(\n",
        "        minimum=test_logged_dataset[\"state\"].min(axis=0),\n",
        "        maximum=test_logged_dataset[\"state\"].max(axis=0),\n",
        "    ),\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 44,
      "metadata": {},
      "outputs": [
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "53cd75a6508c4035adeefce36f7062f6",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "[fit FQE model]:   0%|          | 0/3 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-07-22 18:02:40 [info     ] dataset info                   dataset_info=DatasetInfo(observation_signature=Signature(dtype=[dtype('float64')], shape=[(5,)]), action_signature=Signature(dtype=[dtype('int64')], shape=[(1,)]), reward_signature=Signature(dtype=[dtype('float64')], shape=[(1,)]), action_space=<ActionSpace.DISCRETE: 2>, action_size=5)\n",
            "2023-07-22 18:02:40 [info     ] Directory is created at d3rlpy_logs/DiscreteFQE_20230722180240\n",
            "2023-07-22 18:02:40 [debug    ] Building models...\n",
            "2023-07-22 18:02:40 [debug    ] Models have been built.\n",
            "2023-07-22 18:02:40 [info     ] Parameters                     params={'observation_shape': [5], 'action_size': 5, 'config': {'type': 'fqe', 'params': {'batch_size': 100, 'gamma': 1.0, 'observation_scaler': {'type': 'none', 'params': {}}, 'action_scaler': {'type': 'none', 'params': {}}, 'reward_scaler': {'type': 'none', 'params': {}}, 'learning_rate': 0.0001, 'optim_factory': {'type': 'adam', 'params': {'betas': [0.9, 0.999], 'eps': 1e-08, 'weight_decay': 0, 'amsgrad': False}}, 'encoder_factory': {'type': 'vector', 'params': {'hidden_units': [30, 30], 'activation': 'relu', 'use_batch_norm': False, 'dropout_rate': None, 'use_dense': False, 'exclude_last_activation': False}}, 'q_func_factory': {'type': 'mean', 'params': {'share_encoder': False}}, 'n_critics': 1, 'target_update_interval': 100}}}\n"
          ]
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "0ff8277566dd4494a279da3cb09506e0",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "Epoch 1/1:   0%|          | 0/10000 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-07-22 18:03:05 [info     ] DiscreteFQE_20230722180240: epoch=1 step=10000 epoch=1 metrics={'time_sample_batch': 0.0010205008745193481, 'time_algorithm_update': 0.0013656225442886354, 'loss': 7.594922496168316, 'time_step': 0.002461224675178528} step=10000\n",
            "2023-07-22 18:03:05 [info     ] Model parameters are saved to d3rlpy_logs/DiscreteFQE_20230722180240/model_10000.d3\n",
            "2023-07-22 18:03:05 [info     ] dataset info                   dataset_info=DatasetInfo(observation_signature=Signature(dtype=[dtype('float64')], shape=[(5,)]), action_signature=Signature(dtype=[dtype('int64')], shape=[(1,)]), reward_signature=Signature(dtype=[dtype('float64')], shape=[(1,)]), action_space=<ActionSpace.DISCRETE: 2>, action_size=5)\n",
            "2023-07-22 18:03:05 [info     ] Directory is created at d3rlpy_logs/DiscreteFQE_20230722180305\n",
            "2023-07-22 18:03:05 [debug    ] Building models...\n",
            "2023-07-22 18:03:05 [debug    ] Models have been built.\n",
            "2023-07-22 18:03:05 [info     ] Parameters                     params={'observation_shape': [5], 'action_size': 5, 'config': {'type': 'fqe', 'params': {'batch_size': 100, 'gamma': 1.0, 'observation_scaler': {'type': 'none', 'params': {}}, 'action_scaler': {'type': 'none', 'params': {}}, 'reward_scaler': {'type': 'none', 'params': {}}, 'learning_rate': 0.0001, 'optim_factory': {'type': 'adam', 'params': {'betas': [0.9, 0.999], 'eps': 1e-08, 'weight_decay': 0, 'amsgrad': False}}, 'encoder_factory': {'type': 'vector', 'params': {'hidden_units': [30, 30], 'activation': 'relu', 'use_batch_norm': False, 'dropout_rate': None, 'use_dense': False, 'exclude_last_activation': False}}, 'q_func_factory': {'type': 'mean', 'params': {'share_encoder': False}}, 'n_critics': 1, 'target_update_interval': 100}}}\n"
          ]
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "79dd6a83d48e484681dd04e4c87637d3",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "Epoch 1/1:   0%|          | 0/10000 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-07-22 18:03:27 [info     ] DiscreteFQE_20230722180305: epoch=1 step=10000 epoch=1 metrics={'time_sample_batch': 0.0009497817993164063, 'time_algorithm_update': 0.0011848777770996095, 'loss': 11.54115076702088, 'time_step': 0.0021957416772842406} step=10000\n",
            "2023-07-22 18:03:27 [info     ] Model parameters are saved to d3rlpy_logs/DiscreteFQE_20230722180305/model_10000.d3\n",
            "2023-07-22 18:03:27 [info     ] dataset info                   dataset_info=DatasetInfo(observation_signature=Signature(dtype=[dtype('float64')], shape=[(5,)]), action_signature=Signature(dtype=[dtype('int64')], shape=[(1,)]), reward_signature=Signature(dtype=[dtype('float64')], shape=[(1,)]), action_space=<ActionSpace.DISCRETE: 2>, action_size=5)\n",
            "2023-07-22 18:03:27 [info     ] Directory is created at d3rlpy_logs/DiscreteFQE_20230722180327\n",
            "2023-07-22 18:03:27 [debug    ] Building models...\n",
            "2023-07-22 18:03:27 [debug    ] Models have been built.\n",
            "2023-07-22 18:03:27 [info     ] Parameters                     params={'observation_shape': [5], 'action_size': 5, 'config': {'type': 'fqe', 'params': {'batch_size': 100, 'gamma': 1.0, 'observation_scaler': {'type': 'none', 'params': {}}, 'action_scaler': {'type': 'none', 'params': {}}, 'reward_scaler': {'type': 'none', 'params': {}}, 'learning_rate': 0.0001, 'optim_factory': {'type': 'adam', 'params': {'betas': [0.9, 0.999], 'eps': 1e-08, 'weight_decay': 0, 'amsgrad': False}}, 'encoder_factory': {'type': 'vector', 'params': {'hidden_units': [30, 30], 'activation': 'relu', 'use_batch_norm': False, 'dropout_rate': None, 'use_dense': False, 'exclude_last_activation': False}}, 'q_func_factory': {'type': 'mean', 'params': {'share_encoder': False}}, 'n_critics': 1, 'target_update_interval': 100}}}\n"
          ]
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "2591b7295d3e4acaa8e6ff66fc866df4",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "Epoch 1/1:   0%|          | 0/10000 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "2023-07-22 18:03:47 [info     ] DiscreteFQE_20230722180327: epoch=1 step=10000 epoch=1 metrics={'time_sample_batch': 0.0009308305501937866, 'time_algorithm_update': 0.0009925251245498658, 'loss': 0.12489210550989956, 'time_step': 0.0019779507160186767} step=10000\n",
            "2023-07-22 18:03:47 [info     ] Model parameters are saved to d3rlpy_logs/DiscreteFQE_20230722180327/model_10000.d3\n"
          ]
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "f2eadd7cb1794e28bb9f19d37c9f3e6c",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "[collect input data: eval_policy]:   0%|          | 0/3 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "fecff95ac1ce462f9da7ae05c864b058",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "[calculate on-policy policy value]:   0%|          | 0/100 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "81e008ad24f74668b837271e6fd953cd",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "[calculate on-policy policy value]:   0%|          | 0/100 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "e7af082631a649af975df86654e70527",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "[calculate on-policy policy value]:   0%|          | 0/100 [00:00<?, ?it/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "# takes time\n",
        "input_dict = prep.obtain_whole_inputs(\n",
        "    logged_dataset=test_logged_dataset,\n",
        "    evaluation_policies=evaluation_policies,\n",
        "    require_value_prediction=True,\n",
        "    n_trajectories_on_policy_evaluation=100,\n",
        "    random_state=random_state,\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 45,
      "metadata": {},
      "outputs": [],
      "source": [
        "with open(\"logs/ope_input_dict_discrete.pkl\", \"wb\") as f:\n",
        "    pickle.dump(input_dict, f)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 46,
      "metadata": {},
      "outputs": [],
      "source": [
        "with open(\"logs/ope_input_dict_discrete.pkl\", \"rb\") as f:\n",
        "    input_dict = pickle.load(f)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 49,
      "metadata": {},
      "outputs": [],
      "source": [
        "ope = OPE(\n",
        "    logged_dataset=test_logged_dataset,\n",
        "    ope_estimators=[DM(), TIS(), PDIS(), DR(), SNTIS(), SNPDIS(), SNDR()],\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 50,
      "metadata": {},
      "outputs": [],
      "source": [
        "# performance estimation by ope\n",
        "policy_value_df_dict, policy_value_interval_df_dict = ope.summarize_off_policy_estimates(input_dict, random_state=random_state)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 51,
      "metadata": {},
      "outputs": [
        {
          "data": {
            "text/plain": [
              "{'cql':            policy_value  relative_policy_value\n",
              " on_policy      1.449920               1.000000\n",
              " dm            96.903349              66.833598\n",
              " tis            1.884254               1.299557\n",
              " pdis           1.605809               1.107516\n",
              " dr             7.478347               5.157766\n",
              " sntis          3.176659               2.190921\n",
              " snpdis         1.948656               1.343975\n",
              " sndr         -66.216513             -45.669091,\n",
              " 'ddqn':            policy_value  relative_policy_value\n",
              " on_policy      1.305555               1.000000\n",
              " dm           174.810128             133.897198\n",
              " tis            1.261084               0.965937\n",
              " pdis           1.236058               0.946769\n",
              " dr             3.499134               2.680189\n",
              " sntis          1.247288               0.955370\n",
              " snpdis         1.216990               0.932163\n",
              " sndr           5.413034               4.146156,\n",
              " 'random':            policy_value  relative_policy_value\n",
              " on_policy     -0.475200                    NaN\n",
              " dm            -1.230398                    NaN\n",
              " tis            1.597795                    NaN\n",
              " pdis          -0.054872                    NaN\n",
              " dr            -1.293383                    NaN\n",
              " sntis          0.750280                    NaN\n",
              " snpdis        -0.049983                    NaN\n",
              " sndr          -1.486239                    NaN}"
            ]
          },
          "execution_count": 51,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "# estimated policy performance\n",
        "policy_value_df_dict"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 52,
      "metadata": {},
      "outputs": [
        {
          "data": {
            "text/plain": [
              "{'cql':                 mean  95.0% CI (lower)  95.0% CI (upper)\n",
              " on_policy   1.469558          0.914655          1.948925\n",
              " dm         96.903537         96.483361         97.379692\n",
              " tis         1.889711          1.699092          2.137394\n",
              " pdis        1.600044          1.498639          1.716463\n",
              " dr          7.134107         -9.241690         18.468379\n",
              " sntis       3.185859          2.864494          3.603426\n",
              " snpdis      1.942285          1.814173          2.078125\n",
              " sndr      -66.653928        -94.938943        -48.145270,\n",
              " 'ddqn':                  mean  95.0% CI (lower)  95.0% CI (upper)\n",
              " on_policy    1.334626          0.799851          1.832369\n",
              " dm         174.814123        174.228089        175.471843\n",
              " tis          1.269711          1.056423          1.504950\n",
              " pdis         1.226794          1.115625          1.331387\n",
              " dr           1.623300        -25.687265         22.654888\n",
              " sntis        1.255820          1.044866          1.488485\n",
              " snpdis       1.207907          1.099196          1.310821\n",
              " sndr         3.554284        -23.463584         24.359716,\n",
              " 'random':                mean  95.0% CI (lower)  95.0% CI (upper)\n",
              " on_policy -0.474389         -0.771902         -0.173648\n",
              " dm        -1.230694         -1.238819         -1.221462\n",
              " tis        1.635628         -0.147722          5.240078\n",
              " pdis      -0.072675         -0.389071          0.244077\n",
              " dr        -1.291998         -1.947736         -0.525301\n",
              " sntis      0.768046         -0.069366          2.460596\n",
              " snpdis    -0.066371         -0.357833          0.212516\n",
              " sndr      -1.488357         -1.829563         -1.081294}"
            ]
          },
          "execution_count": 52,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "# estimated confidence intervals of the policy performance\n",
        "policy_value_interval_df_dict"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## 4. Evaluation of OPE methods\n",
        "Now, we evaluate how well the ope estimators estimated the policy value."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 53,
      "metadata": {},
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1400x1200 with 3 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "# we can visualize and compare the result\n",
        "# it seems that DM tends to have a bias though the variance is quite small, \n",
        "# while (trajectory-wise) importance sampling estimators have a smaller bias with a larger variance.\n",
        "# self-normalized estimators also works well\n",
        "ope.visualize_off_policy_estimates(input_dict, random_state=random_state, sharey=False)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 54,
      "metadata": {},
      "outputs": [
        {
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>cql</th>\n",
              "      <th>ddqn</th>\n",
              "      <th>random</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>dm</th>\n",
              "      <td>9111.357085</td>\n",
              "      <td>30103.836849</td>\n",
              "      <td>0.570325</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>tis</th>\n",
              "      <td>0.188646</td>\n",
              "      <td>0.001978</td>\n",
              "      <td>4.297308</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>pdis</th>\n",
              "      <td>0.024301</td>\n",
              "      <td>0.004830</td>\n",
              "      <td>0.176675</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>dr</th>\n",
              "      <td>36.341932</td>\n",
              "      <td>4.811791</td>\n",
              "      <td>0.669424</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>sntis</th>\n",
              "      <td>2.981628</td>\n",
              "      <td>0.003395</td>\n",
              "      <td>1.501802</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>snpdis</th>\n",
              "      <td>0.248738</td>\n",
              "      <td>0.007844</td>\n",
              "      <td>0.180809</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>sndr</th>\n",
              "      <td>4578.746132</td>\n",
              "      <td>16.871384</td>\n",
              "      <td>1.022200</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                cql          ddqn    random\n",
              "dm      9111.357085  30103.836849  0.570325\n",
              "tis        0.188646      0.001978  4.297308\n",
              "pdis       0.024301      0.004830  0.176675\n",
              "dr        36.341932      4.811791  0.669424\n",
              "sntis      2.981628      0.003395  1.501802\n",
              "snpdis     0.248738      0.007844  0.180809\n",
              "sndr    4578.746132     16.871384  1.022200"
            ]
          },
          "execution_count": 54,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "# squared-error (se) measures the estimation accuracy of OPE\n",
        "eval_metric_ope_df = ope.evaluate_performance_of_ope_estimators(\n",
        "    input_dict, \n",
        "    metric=\"se\",\n",
        "    return_by_dataframe=True,\n",
        ")\n",
        "eval_metric_ope_df"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "For more advanced topics in OPE (e.g., cumulative distribution function estimation) and OPS, please refer to [examples/quickstart/rec/rec_synthetic_discrete_advanced.ipynb](https://github.com/hakuhodo-technologies/scope-rl/blob/main/examples/quickstart/rec/rec_synthetic_discrete_advanced.ipynb). \n",
        "\n",
        "For more advanced OPE estimators (e.g., state(-action) marginal estimators and double reinforcement learning), please refer to [examples/quickstart/rec/rec_synthetic_discrete_zoo.ipynb](https://github.com/hakuhodo-technologies/scope-rl/blob/main/examples/quickstart/rec/rec_synthetic_discrete_zoo.ipynb). "
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Reference\n",
        "\n",
        "- Yuta Saito, Shunsuke Aihara, Megumi Matsutani, and Yusuke Narita. \\\n",
        "\"Open Bandit Dataset and Pipeline: Towards Realistic and Reproducible Off-Policy Evaluation.\", 2021.\n",
        "\n",
        "- Takuma Seno and Michita Imai. \\\n",
        "\"d3rlpy: An Offline Deep Reinforcement Library.\", 2021.\n",
        "\n",
        "- Sergey Levine, Aviral Kumar, George Tucker, and Justin Fu. \\\n",
        "\"Offline Reinforcement Learning: Tutorial, Review, and Perspectives on Open Problems.\" 2020.\n",
        "\n",
        "- Aviral Kumar, Aurick Zhou, George Tucker, and Sergey Levine. \\\n",
        "\"Conservative Q-Learning for Offline Reinforcement Learning.\", 2020.\n",
        "\n",
        "- Nathan Kallus and Masatoshi Uehara. \\\n",
        "\"Intrinsically Efficient, Stable, and Bounded Off-Policy Evaluation for Reinforcement Learning.\", 2019.\n",
        "\n",
        "- Hoang Le, Cameron Voloshin, and Yisong Yue. \\\n",
        "\"Batch Policy Learning under Constraints.\", 2019.\n",
        "\n",
        "- Di Wu, Xiujun Chen, Xun Yang, Hao Wang, Qing Tan, Xiaoxun Zhang, Jian Xu, and Kun Gai. \\\n",
        "\"Budget Constrained Bidding by Model-free Reinforcement Learning in Display Advertising.\", 2018.\n",
        "\n",
        "- Jun Zhao, Guang Qiu, Ziyu Guan, Wei Zhao, and Xiaofei He. \\\n",
        "\"Deep Reinforcement Learning for Sponsored Search Real-time Bidding.\", 2018.\n",
        "\n",
        "- Nan Jiang and Lihong Li. \\\n",
        "\"Doubly Robust Off-policy Value Evaluation for Reinforcement Learning.\", 2016.\n",
        "\n",
        "- Philip S. Thomas and Emma Brunskill. \\\n",
        "\"Data-Efficient Off-Policy Policy Evaluation for Reinforcement Learning.\", 2016.\n",
        "\n",
        "- Greg Brockman, Vicki Cheung, Ludwig Pettersson, Jonas Schneider, John Schulman, Jie Tang, and Wojciech Zaremba. \\\n",
        "\"OpenAI Gym.\", 2016.\n",
        "\n",
        "- Adith Swaminathan and Thorsten Joachims. \\\n",
        "\"The Self-Normalized Estimator for Counterfactual Learning.\", 2015.\n",
        "\n",
        "- Hado van Hasselt, Arthur Guez, and David Silver. \\\n",
        "\"Deep Reinforcement Learning with Double Q-learning.\", 2015.\n",
        "\n",
        "- Miroslav Dudík, Dumitru Erhan, John Langford, and Lihong Li. \\\n",
        "\"Doubly Robust Policy Evaluation and Optimization.\", 2014.\n",
        "\n",
        "- Alex Strehl, John Langford, Sham Kakade, and Lihong Li. \\\n",
        "\"Learning from Logged Implicit Exploration Data.\", 2010.\n",
        "\n",
        "- Alina Beygelzimer and John Langford. \\\n",
        "\"The Offset Tree for Learning with Partial Labels.\", 2009.\n",
        "\n",
        "- Doina Precup, Richard S. Sutton, and Satinder P. Singh. \\\n",
        "\"Eligibility Traces for Off-Policy Policy Evaluation.\", 2000."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": []
    }
  ],
  "metadata": {
    "kernelspec": {
      "display_name": "Python 3 (ipykernel)",
      "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": "3.9.6"
    },
    "vscode": {
      "interpreter": {
        "hash": "70404ee114725fce8ed9e697d67827f8546c678889944e6d695790702cbfe1f5"
      }
    }
  },
  "nbformat": 4,
  "nbformat_minor": 4
}