Copy_of_Modelling_Cognitive_Tasks.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "execution": {},
        "id": "Azi_bLWn6mAR"
      },
      "source": [
        "# Using RL to Model Cognitive Tasks\n",
        "\n",
        "**By Neurmatch Academy**\n",
        "\n",
        "__Content creators:__ Morteza Ansarinia, Yamil Vidal\n",
        "\n",
        "__Production editor:__ Spiros Chavlis\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "execution": {},
        "id": "t0JJoEeC6mAb"
      },
      "source": [
        "<p align='center'><img src='https://github.com/NeuromatchAcademy/widgets/blob/master/sponsors.png?raw=True'/></p>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "execution": {},
        "id": "9HbuVnWO6mAc"
      },
      "source": [
        "---\n",
        "# Objective\n",
        "\n",
        "- This project aims to use behavioral data to train an agent and then use the agent to investigate data produced by human subjects. Having a computational agent that mimics humans in such tests, we will be able to compare its mechanics with human data.\n",
        "\n",
        "- In another conception, we could fit an agent that learns many cognitive tasks that require abstract-level constructs such as executive functions. This is a multi-task control problem.\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "execution": {},
        "id": "-hzSA9mE6mAd"
      },
      "source": [
        "---\n",
        "# Setup"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "execution": {},
        "id": "zigNT_hI6mAd",
        "outputId": "2d7fc1e1-12bb-4155-d7f9-536a2cda7b4b",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[?25l\r\u001b[K     |█                               | 10 kB 30.0 MB/s eta 0:00:01\r\u001b[K     |██                              | 20 kB 20.5 MB/s eta 0:00:01\r\u001b[K     |███▏                            | 30 kB 10.6 MB/s eta 0:00:01\r\u001b[K     |████▏                           | 40 kB 4.4 MB/s eta 0:00:01\r\u001b[K     |█████▏                          | 51 kB 4.3 MB/s eta 0:00:01\r\u001b[K     |██████▎                         | 61 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |███████▎                        | 71 kB 5.5 MB/s eta 0:00:01\r\u001b[K     |████████▍                       | 81 kB 5.5 MB/s eta 0:00:01\r\u001b[K     |█████████▍                      | 92 kB 6.1 MB/s eta 0:00:01\r\u001b[K     |██████████▍                     | 102 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |███████████▌                    | 112 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |████████████▌                   | 122 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |█████████████▋                  | 133 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |██████████████▋                 | 143 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |███████████████▋                | 153 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |████████████████▊               | 163 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |█████████████████▊              | 174 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |██████████████████▉             | 184 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |███████████████████▉            | 194 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |████████████████████▉           | 204 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |██████████████████████          | 215 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |███████████████████████         | 225 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |████████████████████████        | 235 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |█████████████████████████       | 245 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |██████████████████████████      | 256 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |███████████████████████████▏    | 266 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |████████████████████████████▏   | 276 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |█████████████████████████████▏  | 286 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |██████████████████████████████▎ | 296 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |███████████████████████████████▎| 307 kB 5.1 MB/s eta 0:00:01\r\u001b[K     |████████████████████████████████| 314 kB 5.1 MB/s \n",
            "\u001b[33mWARNING: dm-acme 0.4.0 does not provide the extra 'tensorflow'\u001b[0m\n",
            "\u001b[K     |████████████████████████████████| 3.8 MB 45.2 MB/s \n",
            "\u001b[K     |████████████████████████████████| 72 kB 676 kB/s \n",
            "\u001b[K     |████████████████████████████████| 342 kB 62.4 MB/s \n",
            "\u001b[K     |████████████████████████████████| 197 kB 73.3 MB/s \n",
            "\u001b[K     |████████████████████████████████| 145 kB 72.8 MB/s \n",
            "\u001b[K     |████████████████████████████████| 117 kB 77.4 MB/s \n",
            "\u001b[K     |████████████████████████████████| 6.4 MB 23.9 MB/s \n",
            "\u001b[K     |████████████████████████████████| 4.0 MB 46.9 MB/s \n",
            "\u001b[K     |████████████████████████████████| 668.3 MB 18 kB/s \n",
            "\u001b[K     |████████████████████████████████| 5.7 MB 60.7 MB/s \n",
            "\u001b[K     |████████████████████████████████| 462 kB 75.4 MB/s \n",
            "\u001b[K     |████████████████████████████████| 217 kB 70.6 MB/s \n",
            "\u001b[K     |████████████████████████████████| 596 kB 64.6 MB/s \n",
            "\u001b[K     |████████████████████████████████| 51 kB 7.2 MB/s \n",
            "\u001b[K     |████████████████████████████████| 272 kB 73.0 MB/s \n",
            "\u001b[?25h  Building wheel for dm-acme (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "\u001b[K     |████████████████████████████████| 18.8 MB 5.2 MB/s \n",
            "\u001b[K     |████████████████████████████████| 88 kB 8.5 MB/s \n",
            "\u001b[K     |████████████████████████████████| 4.9 MB 47.1 MB/s \n",
            "\u001b[K     |████████████████████████████████| 207 kB 67.9 MB/s \n",
            "\u001b[K     |████████████████████████████████| 99 kB 9.5 MB/s \n",
            "\u001b[?25h  Building wheel for bsuite (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "\u001b[K     |████████████████████████████████| 254 kB 5.1 MB/s \n",
            "\u001b[K     |████████████████████████████████| 104 kB 5.3 MB/s \n",
            "\u001b[?25h"
          ]
        }
      ],
      "source": [
        "# @title Install dependencies\n",
        "# !pip install --upgrade pip setuptools wheel --quiet\n",
        "!pip install dm-acme[jax,tensorflow] --quiet\n",
        "!pip install dm-acme[envs] --quiet\n",
        "!pip install dm-sonnet --quiet\n",
        "!pip install trfl --quiet"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "execution": {},
        "id": "IOdC2K-56mAk"
      },
      "outputs": [],
      "source": [
        "# Imports\n",
        "\n",
        "import time\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import sonnet as snt\n",
        "import seaborn as sns\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "import dm_env\n",
        "\n",
        "import acme\n",
        "from acme import specs\n",
        "from acme import wrappers\n",
        "from acme import EnvironmentLoop\n",
        "from acme.agents.tf import dqn\n",
        "from acme.utils import loggers\n",
        "from acme.agents.tf import r2d2\n",
        "from tqdm import tqdm"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "execution": {},
        "id": "VIvUNa5t6mAl"
      },
      "outputs": [],
      "source": [
        "# @title Figure settings\n",
        "from IPython.display import clear_output, display, HTML\n",
        "%matplotlib inline\n",
        "sns.set()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "execution": {},
        "id": "-R0grYI-6mAm"
      },
      "source": [
        "---\n",
        "# Background\n",
        "\n",
        "- Cognitive scientists use standard lab tests to tap into specific processes in the brain and behavior. Some examples of those tests are Stroop, N-back, Digit Span, TMT (Trail making tests), and WCST (Wisconsin Card Sorting Tests).\n",
        "\n",
        "- Despite an extensive body of research that explains human performance using descriptive what-models, we still need a more sophisticated approach to gain a better understanding of the underlying processes (i.e., a how-model).\n",
        "\n",
        "- Interestingly, many of such tests can be thought of as a continuous stream of stimuli and corresponding actions, that is in consonant with the RL formulation. In fact, RL itself is in part motivated by how the brain enables goal-directed behaviors using reward systems, making it a good choice to explain human performance.\n",
        "\n",
        "- One behavioral test example would be the N-back task.\n",
        "\n",
        "  - In the N-back, participants view a sequence of stimuli, one by one, and are asked to categorize each stimulus as being either match or non-match. Stimuli are usually numbers, and feedback is given at both timestep and trajectory levels.\n",
        "\n",
        "  - The agent is rewarded when its response matches the stimulus that was shown N steps back in the episode. A simpler version of the N-back uses two-choice action schema, that is match vs non-match. Once the present stimulus matches the one presented N step back, then the agent is expected to respond to it as being a `match`.\n",
        "\n",
        "\n",
        "- Given a trained RL agent, we then find correlates of its fitted parameters with the brain mechanisms. The most straightforward composition could be the correlation of model parameters with the brain activities."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "execution": {},
        "id": "Jr2eTAmn6mAo"
      },
      "source": [
        "## Datasets\n",
        "\n",
        "- HCP WM task ([NMA-CN HCP notebooks](https://github.com/NeuromatchAcademy/course-content/tree/master/projects/fMRI))\n",
        "\n",
        "Any dataset that used cognitive tests would work.\n",
        "Question: limit to behavioral data vs fMRI?\n",
        "Question: Which stimuli and actions to use?\n",
        "classic tests can be modeled using 1) bounded symbolic stimuli/actions (e.g., A, B, C), but more sophisticated one would require texts or images (e.g., face vs neutral images in social stroop dataset)\n",
        "The HCP dataset from NMA-CN contains behavioral and imaging data for 7 cognitive tests including various versions of N-back."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "execution": {},
        "id": "hi23aJh46mAp"
      },
      "source": [
        "## N-back task\n",
        "\n",
        "In the N-back task, participants view a sequence of stimuli, one per time, and are asked to categorize each stimulus as being either match or non-match. Stimuli are usually numbers, and feedbacks are given at both timestep and trajectory levels.\n",
        "\n",
        "In a typical neuro setup, both accuracy and response time are measured, but here, for the sake of brevity, we focus only on accuracy of responses."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "execution": {},
        "id": "0BCw1SqR6mAq"
      },
      "source": [
        "---\n",
        "# Cognitive Tests Environment\n",
        "\n",
        "First we develop an environment in that agents perform a cognitive test, here the N-back."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "execution": {},
        "id": "H7aDZqad6mAr"
      },
      "source": [
        "## Human dataset\n",
        "\n",
        "We need a dataset of human perfoming a N-back test, with the following features:\n",
        "\n",
        "- `participant_id`: following the BIDS format, it contains a unique identifier for each participant.\n",
        "- `trial_index`: same as `time_step`.\n",
        "- `stimulus`: same as `observation`.\n",
        "- `response`: same as `action`, recorded response by the human subject.\n",
        "- `expected_response`: correct response.\n",
        "- `is_correct`: same as `reward`, whether the human subject responded correctly.\n",
        "- `response_time`: won't be used here.\n",
        "\n",
        "Here we generate a mock dataset with those features, but remember to **replace this with real human data.**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "execution": {},
        "id": "mvgbdQuK6mAs"
      },
      "source": [
        "### Environment\n",
        "\n",
        "The following cell implments N-back envinronment, that we later use to train a RL agent on human data. It is capable of performing two kinds of simulation:\n",
        "- rewards the agent once the action was correct (i.e., a normative model of the environment).\n",
        "- receives human data (or mock data if you prefer), and returns what participants performed as the observation. This is more useful for preference-based RL."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "execution": {},
        "id": "ufg4-znW6mAs"
      },
      "outputs": [],
      "source": [
        "class NBack(dm_env.Environment):\n",
        "\n",
        "  ACTIONS = ['match', 'non-match']\n",
        "\n",
        "  def __init__(self,\n",
        "               N=2,\n",
        "               episode_steps=32,\n",
        "               stimuli_choices=list('ABCDEF'),\n",
        "               human_data=None,\n",
        "               seed=1,\n",
        "               prob = None\n",
        "               ):\n",
        "    \"\"\"\n",
        "    Args:\n",
        "      N: Number of steps to look back for the matched stimuli. Defaults to 2 (as in 2-back).\n",
        "      episode_steps\n",
        "      stimuli_choices\n",
        "      human_data\n",
        "      seed\n",
        "\n",
        "    \"\"\"\n",
        "    self.N = N\n",
        "    self.episode_steps = episode_steps\n",
        "    self.stimuli_choices = stimuli_choices\n",
        "    self.stimuli = np.empty(shape=episode_steps)  # will be filled in the `reset()`\n",
        "    self.n_s = len(self.stimuli_choices)\n",
        "    if prob is None:\n",
        "      self.prob = 0.5 * (self.n_s - 2)/ (self.n_s - 1)\n",
        "    else:\n",
        "      self.prob = prob\n",
        "    self._reset_next_step = True\n",
        "\n",
        "    # whether mimic humans or reward the agent once it responds optimally.\n",
        "    if human_data is None:\n",
        "      self._imitate_human = False\n",
        "      self.human_data = None\n",
        "      self.human_subject_data = None\n",
        "    else:\n",
        "      self._imitate_human = True\n",
        "      self.human_data = human_data\n",
        "      self.human_subject_data = None\n",
        "\n",
        "    self._action_history = []\n",
        "\n",
        "  def reset(self):\n",
        "    self._reset_next_step = False\n",
        "    self._current_step = 0\n",
        "    self._action_history.clear()\n",
        "\n",
        "    # generate a random sequence instead of relying on human data\n",
        "    if self.human_data is None:\n",
        "      for i in range(self.episode_steps):\n",
        "          if i<self.N:\n",
        "            self.stimuli[i]=np.random.choice(self.n_s)\n",
        "          else:\n",
        "            rand = np.random.random()\n",
        "            if rand < self.prob:\n",
        "              self.stimuli[i]=self.stimuli[i-self.N]\n",
        "            else:\n",
        "              self.stimuli[i]=np.random.choice(self.n_s)\n",
        "    else:\n",
        "      self.human_subject_data = self.human_data.query('participant_id == participant_id.sample().iloc[0]',\n",
        "                                                engine='python').sort_values('trial_index')\n",
        "      self.stimuli = self.human_subject_data['stimulus'].to_list()\n",
        "      self.stimuli = np.array([ord(s) - ord('A') + 1 for s in self.stimuli]).astype(np.float32)\n",
        "\n",
        "    return dm_env.restart(self._observation())\n",
        "\n",
        "\n",
        "  def _episode_return(self):\n",
        "    if self._imitate_human:\n",
        "      return np.mean(self.human_subject_data['response'] == self._action_history)\n",
        "    else:\n",
        "      return 0.0\n",
        "\n",
        "  def step(self, action: int):\n",
        "    if self._reset_next_step:\n",
        "      return self.reset()\n",
        "\n",
        "    agent_action = NBack.ACTIONS[action]\n",
        "\n",
        "    if self._imitate_human:\n",
        "      human_action = self.human_subject_data['response'].iloc[self._current_step]\n",
        "      step_reward = 0. if (agent_action == human_action) else -1.\n",
        "    else:\n",
        "      expected_action = 'match' if (self.stimuli[self._current_step] == self.stimuli[self._current_step - self.N]) else 'non-match'\n",
        "      step_reward = 0. if (agent_action == expected_action) else -1.\n",
        "\n",
        "    self._action_history.append(agent_action)\n",
        "\n",
        "    self._current_step += 1\n",
        "\n",
        "    if self._current_step == self.stimuli.shape[0]:\n",
        "      self._reset_next_step = True\n",
        "      return dm_env.termination(reward=self._episode_return(),\n",
        "                                observation=self._observation())\n",
        "    else:\n",
        "      return dm_env.transition(reward=step_reward,\n",
        "                               observation=self._observation())\n",
        "\n",
        "  def observation_spec(self):\n",
        "    return dm_env.specs.BoundedArray(\n",
        "        shape=self.stimuli.shape,\n",
        "        dtype=self.stimuli.dtype,\n",
        "        name='nback_stimuli', minimum=0, maximum=len(self.stimuli_choices) + 1)\n",
        "\n",
        "  def action_spec(self):\n",
        "    return dm_env.specs.DiscreteArray(\n",
        "        num_values=len(NBack.ACTIONS),\n",
        "        dtype=np.int32,\n",
        "        name='action')\n",
        "\n",
        "  def _observation(self):\n",
        "    obs = self.stimuli[self._current_step - 1]\n",
        "    # obs = self.stimuli[:self._current_step+1].copy()\n",
        "    # obs = np.pad(obs,(0, len(self.stimuli) - len(obs)))\n",
        "\n",
        "    return obs\n",
        "\n",
        "  def plot_state(self):\n",
        "    \"\"\"Display current state of the environment.\n",
        "\n",
        "     Note: `M` mean `match`, and `.` is a `non-match`.\n",
        "    \"\"\"\n",
        "    stimuli = self.stimuli[:self._current_step - 1]\n",
        "    actions = ['M' if a=='match' else '.' for a in self._action_history[:self._current_step - 1]]\n",
        "    return HTML(\n",
        "        f'<b>Environment ({self.N}-back):</b><br />'\n",
        "        f'<pre><b>Stimuli:</b> {\"\".join(map(str,map(int,stimuli)))}</pre>'\n",
        "        f'<pre><b>Actions:</b> {\"\".join(actions)}</pre>'\n",
        "    )\n",
        "\n",
        "  @staticmethod\n",
        "  def create_environment(prob = None):\n",
        "    \"\"\"Utility function to create a N-back environment and its spec.\"\"\"\n",
        "\n",
        "    # Make sure the environment outputs single-precision floats.\n",
        "    environment = wrappers.SinglePrecisionWrapper(NBack(prob = prob))\n",
        "\n",
        "    # Grab the spec of the environment.\n",
        "    environment_spec = specs.make_environment_spec(environment)\n",
        "\n",
        "    return environment, environment_spec"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "execution": {},
        "id": "NQTnbj5V6mAt"
      },
      "source": [
        "### Define a random agent\n",
        "\n",
        "For more information you can refer to NMA-DL W3D2 Basic Reinforcement learning."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "execution": {},
        "id": "_h1bWf8D6mAu"
      },
      "outputs": [],
      "source": [
        "class RandomAgent(acme.Actor):\n",
        "\n",
        "  def __init__(self, environment_spec, action = None):\n",
        "    \"\"\"Gets the number of available actions from the environment spec.\"\"\"\n",
        "    self._num_actions = environment_spec.actions.num_values\n",
        "    self.action = action\n",
        "\n",
        "  def select_action(self, observation):\n",
        "    \"\"\"Selects an action uniformly at random.\"\"\"\n",
        "    if self.action is not None: \n",
        "      return self.action\n",
        "    return np.random.randint(self._num_actions)\n",
        "\n",
        "  def observe_first(self, timestep):\n",
        "    \"\"\"Does not record as the RandomAgent has no use for data.\"\"\"\n",
        "    pass\n",
        "\n",
        "  def observe(self, action, next_timestep):\n",
        "    \"\"\"Does not record as the RandomAgent has no use for data.\"\"\"\n",
        "    pass\n",
        "\n",
        "  def update(self):\n",
        "    \"\"\"Does not update as the RandomAgent does not learn from data.\"\"\"\n",
        "    pass"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Define Learnable Agents"
      ],
      "metadata": {
        "id": "_X_RIcItvaVU"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def dqn_make_network(action_spec: specs.DiscreteArray) -> snt.Module:\n",
        "  return snt.Sequential([\n",
        "      snt.Flatten(),\n",
        "      snt.nets.MLP([128, 128, action_spec.num_values]),\n",
        "  ])\n",
        "\n",
        "def r2d2_make_network(action_spec: specs.DiscreteArray) -> snt.Module:\n",
        "  return snt.DeepRNN([\n",
        "            snt.Flatten(),\n",
        "            snt.LSTM(hidden_size = 10),\n",
        "            snt.nets.MLP([128, 128, action_spec.num_values])\n",
        "            ])"
      ],
      "metadata": {
        "id": "T6A2pa9Zvcpz"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "execution": {},
        "id": "135ZABGG6mAw"
      },
      "source": [
        "### Get reward distribution and mean reward of the agent"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "execution": {},
        "id": "VPawgcGr6mAx"
      },
      "outputs": [],
      "source": [
        "def get_reward_distribution(agent, env, n_episodes = 1000,  episode_length = 32, log_loss = False):\n",
        "  # main loop\n",
        "  n_total_steps = 0\n",
        "  n_steps = n_episodes * episode_length\n",
        "  all_returns = []\n",
        "  for episode in range(n_episodes):\n",
        "    episode_steps = 0\n",
        "    episode_return = 0\n",
        "    episode_loss = 0\n",
        "\n",
        "    start_time = time.time()\n",
        "\n",
        "    timestep = env.reset()\n",
        "\n",
        "    # Make the first observation.\n",
        "    agent.observe_first(timestep)\n",
        "\n",
        "    # Run an episode\n",
        "    while not timestep.last():\n",
        "\n",
        "      # Generate an action from the agent's policy and step the environment.\n",
        "      action = agent.select_action(timestep.observation)\n",
        "      timestep = env.step(action)\n",
        "\n",
        "      # Have the agent observe the timestep and let the agent update itself.\n",
        "      agent.observe(action, next_timestep=timestep)\n",
        "      agent.update()\n",
        "\n",
        "      # Book-keeping.\n",
        "      episode_steps += 1\n",
        "      n_total_steps += 1\n",
        "      episode_return += timestep.reward\n",
        "\n",
        "      if log_loss:\n",
        "        episode_loss += agent.last_loss\n",
        "\n",
        "      if n_steps is not None and n_total_steps >= n_steps:\n",
        "        break\n",
        "\n",
        "    # Collect the results and combine with counts.\n",
        "    steps_per_second = episode_steps / (time.time() - start_time)\n",
        "    result = {\n",
        "        'episode': episode,\n",
        "        'episode_length': episode_steps,\n",
        "        'episode_return': episode_return,\n",
        "    }\n",
        "    if log_loss:\n",
        "      result['loss_avg'] = episode_loss/episode_steps\n",
        "\n",
        "    all_returns.append(episode_return)\n",
        "\n",
        "    display(env.plot_state())\n",
        "    # Log the given results.\n",
        "    # print(result)\n",
        "\n",
        "    if n_steps is not None and n_total_steps >= n_steps:\n",
        "      break\n",
        "\n",
        "  return all_returns, np.mean(all_returns), np.std(all_returns)"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Training and Evaluation Script"
      ],
      "metadata": {
        "id": "DMxo_yt5uN5d"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def train_and_eval(env, agent, train = False, n_episodes = 1000,  episode_length = 32, log_loss = False):\n",
        "  if train:\n",
        "    loop = EnvironmentLoop(env, agent, logger=loggers.InMemoryLogger())\n",
        "    loop.run(n_episodes * 100)\n",
        "\n",
        "  all_returns, mean_return, std_return = get_reward_distribution(agent, env, n_episodes = 1000,  episode_length = 32, log_loss = False)\n",
        "  clear_output()\n",
        "  print(f'Mean Return: {mean_return} | Standard Deviation: {std_return}')\n",
        "  plt.figure()\n",
        "  sns.histplot(all_returns, stat=\"density\", kde=True, bins=12)\n",
        "  plt.xlabel('Return [a.u.]')\n",
        "  plt.ylabel('Density')\n",
        "  plt.show()\n",
        "\n",
        "  if train:\n",
        "    return pd.DataFrame(loop._logger._data)\n",
        "  else:\n",
        "    return None"
      ],
      "metadata": {
        "id": "_cxysHVguQOG"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "execution": {},
        "id": "_EGCq1_c6mAu"
      },
      "source": [
        "### Initialize the environments"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "execution": {},
        "id": "tuD4BchT6mAv",
        "outputId": "18fb389b-5d40-4acb-ad57-5a2beb450899",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "actions:\n",
            " DiscreteArray(shape=(), dtype=int32, name=action, minimum=0, maximum=1, num_values=2)\n",
            "observations:\n",
            " BoundedArray(shape=(32,), dtype=dtype('float32'), name='nback_stimuli', minimum=0.0, maximum=7.0)\n",
            "rewards:\n",
            " Array(shape=(), dtype=dtype('float32'), name='reward')\n"
          ]
        }
      ],
      "source": [
        "env_balanced, env_spec_balanced = NBack.create_environment()\n",
        "env_unbalanced, env_spec_unbalanced = NBack.create_environment(prob = 0.)\n",
        "\n",
        "print('actions:\\n', env_spec_balanced.actions)\n",
        "print('observations:\\n', env_spec_balanced.observations)\n",
        "print('rewards:\\n', env_spec_balanced.rewards)"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Define Agents"
      ],
      "metadata": {
        "id": "ukAvPj94hwOJ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "random_agent = RandomAgent(env_spec_balanced)\n",
        "match_agent = RandomAgent(env_spec_balanced, 0)\n",
        "unmatch_agent = RandomAgent(env_spec_balanced, 1)\n",
        "\n",
        "dqn_agent_balanced = dqn.DQN(\n",
        "    environment_spec=env_spec_balanced,\n",
        "    network=dqn_make_network(env_spec_balanced.actions),\n",
        "    epsilon=[0.5],\n",
        "    learning_rate = 0.0001,\n",
        "    min_replay_size = 32,\n",
        "    logger=loggers.InMemoryLogger(),\n",
        "    checkpoint=False,\n",
        ")\n",
        "\n",
        "dqn_agent_unbalanced = dqn.DQN(\n",
        "    environment_spec=env_spec_unbalanced,\n",
        "    network=dqn_make_network(env_spec_unbalanced.actions),\n",
        "    epsilon=[0.5],\n",
        "    learning_rate = 0.0001,\n",
        "    min_replay_size = 32,\n",
        "    logger=loggers.InMemoryLogger(),\n",
        "    checkpoint=False,\n",
        ")\n",
        "\n",
        "drqn_agent_balanced = r2d2.R2D2(\n",
        "    environment_spec=env_spec_balanced,\n",
        "    network=r2d2_make_network(env_spec_balanced.actions),\n",
        "    learning_rate = 0.0001,\n",
        "    burn_in_length=10,\n",
        "    trace_length=2,\n",
        "    replay_period=10,\n",
        "    epsilon=[0.05],\n",
        "    logger=loggers.InMemoryLogger(),\n",
        "    min_replay_size = 32,\n",
        "    checkpoint=False,\n",
        ")\n",
        "\n",
        "drqn_agent_unbalanced = r2d2.R2D2(\n",
        "    environment_spec=env_spec_unbalanced,\n",
        "    network=r2d2_make_network(env_spec_unbalanced.actions),\n",
        "    learning_rate = 0.0001,\n",
        "    burn_in_length=10,\n",
        "    trace_length=2,\n",
        "    replay_period=10,\n",
        "    epsilon=[0.05],\n",
        "    logger=loggers.InMemoryLogger(),\n",
        "    min_replay_size = 32,\n",
        "    checkpoint=False,\n",
        ")"
      ],
      "metadata": {
        "id": "XcvCnGfDhxIC"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "n_episodes = 250\n",
        "episode_length = 32"
      ],
      "metadata": {
        "id": "x4yOQ6aY6P9K"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Random Agent"
      ],
      "metadata": {
        "id": "tI_K_1zDyaz7"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Balanced N-Back"
      ],
      "metadata": {
        "id": "SURbpHPeydLy"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "train_and_eval(env_balanced, random_agent, n_episodes = n_episodes,  episode_length = episode_length, log_loss = False)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 302
        },
        "id": "7KqHQnnXw-oX",
        "outputId": "b204127f-2b72-49e9-bea2-ee568f3fe196"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mean Return: -15.498 | Standard Deviation: 2.7626067400192884\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Un-Balanced N-Back"
      ],
      "metadata": {
        "id": "sPQgEENVyf8O"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "train_and_eval(env_unbalanced, random_agent, n_episodes = n_episodes,  episode_length = episode_length, log_loss = False)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 302
        },
        "id": "4eDnhTWYyBbM",
        "outputId": "b7fe34a4-ca61-483e-f6c4-3cd4d8521632"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mean Return: -15.543 | Standard Deviation: 2.8000269641558813\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Matching Agent"
      ],
      "metadata": {
        "id": "KHlNxPbMyhzS"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Balanced N-Back"
      ],
      "metadata": {
        "id": "uhblKbMfyjmh"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "train_and_eval(env_balanced, match_agent, n_episodes = n_episodes,  episode_length = episode_length, log_loss = False)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 303
        },
        "id": "iSwcN1cHyEJf",
        "outputId": "e4507f77-5b39-4228-a50f-e4ab1b893ce5"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mean Return: -16.112 | Standard Deviation: 2.7479912663616672\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Un-Balanced N-Back"
      ],
      "metadata": {
        "id": "JO5ZCYnHylS0"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "train_and_eval(env_unbalanced, match_agent, n_episodes = n_episodes,  episode_length = episode_length, log_loss = False)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 302
        },
        "id": "AfdlslUeyEg3",
        "outputId": "67ddcc79-4ff5-4767-8ad4-5a43083eae0a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mean Return: -25.856 | Standard Deviation: 2.0467691613858165\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Un-Matching Agent"
      ],
      "metadata": {
        "id": "v9KqkvKRyoO4"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Balanced N-Back"
      ],
      "metadata": {
        "id": "-jRSzfu2yqQP"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "train_and_eval(env_balanced, unmatch_agent, n_episodes = n_episodes,  episode_length = episode_length, log_loss = False)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 302
        },
        "id": "CF2AnpQiyJxl",
        "outputId": "950f0470-4cb3-483a-bcd1-96172ee0df3f"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mean Return: -14.956 | Standard Deviation: 2.766597910792242\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Un-Balanced N-Back"
      ],
      "metadata": {
        "id": "lziRjJDYyrD2"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "train_and_eval(env_unbalanced, unmatch_agent, n_episodes = n_episodes,  episode_length = episode_length, log_loss = False)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 302
        },
        "id": "eULQxh4KyKGB",
        "outputId": "70e13c32-0cc7-461c-ad69-d70d6fc3719c"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mean Return: -5.231 | Standard Deviation: 2.125944260793307\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## DQN Agent"
      ],
      "metadata": {
        "id": "ji6I5b0zys59"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Balanced N-Back"
      ],
      "metadata": {
        "id": "1tCSFK6RyuPb"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "logs_dqn_balanced = train_and_eval(env_balanced, dqn_agent_balanced, True, n_episodes = n_episodes,  episode_length = episode_length, log_loss = False)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 302
        },
        "id": "TR4-WWlQyOAc",
        "outputId": "46b332aa-9e2c-4586-92ed-8e32908f31e1"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mean Return: -13.815 | Standard Deviation: 2.7554990473596614\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Un-Balanced N-Back"
      ],
      "metadata": {
        "id": "JRApxz2xyvIn"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "logs_dqn_unbalanced = train_and_eval(env_unbalanced, dqn_agent_unbalanced, True, n_episodes = n_episodes,  episode_length = episode_length, log_loss = False)"
      ],
      "metadata": {
        "id": "RbHSBs9JyOPL",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 302
        },
        "outputId": "90d92d01-889c-4162-ca8c-080c3a98f771"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mean Return: -10.425 | Standard Deviation: 2.656383820158525\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## DRQN Agent"
      ],
      "metadata": {
        "id": "VYENVrXxywgf"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Balanced N-Back"
      ],
      "metadata": {
        "id": "PI2ZXu7Kyxyc"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "logs_drqn_balanced = train_and_eval(env_balanced, drqn_agent_balanced, True, n_episodes = n_episodes,  episode_length = episode_length, log_loss = False)"
      ],
      "metadata": {
        "id": "FKEeEbjByWrp",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "397476f4-3869-4125-9218-fbce6ba349a9"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "INFO:tensorflow:Assets written to: /root/acme/b2be4e62-0c1f-11ed-8b42-0242ac1c0002/snapshots/network/assets\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Un-Balanced N-Back"
      ],
      "metadata": {
        "id": "h7hLLNggyyRG"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "logs_drqn_unbalanced = train_and_eval(env_unbalanced, drqn_agent_unbalanced, True, n_episodes = n_episodes,  episode_length = episode_length, log_loss = False)"
      ],
      "metadata": {
        "id": "y0Kz9jPtyW78"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def moving_average(interval, window_size):\n",
        "    window = np.ones(int(window_size)) / float(window_size)\n",
        "    return np.convolve(interval, window, 'same')  "
      ],
      "metadata": {
        "id": "uro52jc69oCH"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Balanced Logs\n",
        "dqn_b_ma = moving_average(logs_dqn_balanced['episode_return'], 100)\n",
        "drqn_b_ma = moving_average(logs_drqn_balanced['episode_return'], 100)\n",
        "\n",
        "# dqn_b = list(logs_dqn_balanced['episode_return'])\n",
        "# drqn_b = list(logs_drqn_balanced['episode_return'])\n",
        "\n",
        "plt.plot(dqn_b_ma, color='blue')\n",
        "plt.plot(drqn_b_ma, color='yellow')\n",
        "\n",
        "# plt.plot(dqn_b, color='blue')\n",
        "# plt.plot(drqn_b, color='yellow')\n",
        "\n",
        "plt.ylabel(\"episode_reward\")\n",
        "plt.xlabel(\"episode (N_steps per episode = 32)\")\n",
        "plt.legend(['DQN','DRQN'])\n",
        "plt.title(\"Equal match and non-match\")"
      ],
      "metadata": {
        "id": "gLIG5S-K1tB-"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Un-Balanced Logs\n",
        "dqn_u_ma = moving_average(logs_dqn_unbalanced['episode_return'], 100)\n",
        "drqn_u_ma = moving_average(logs_drqn_unbalanced['episode_return'], 100)\n",
        "\n",
        "# dqn_u = list(logs_dqn_unbalanced['episode_return'])\n",
        "# drqn_u = list(logs_drqn_unbalanced['episode_return'])\n",
        "\n",
        "plt.plot(dqn_u_ma, color='blue')\n",
        "plt.plot(drqn_u_ma, color='yellow')\n",
        "\n",
        "# plt.plot(dqn_u, color='blue')\n",
        "# plt.plot(drqn_u, color='yellow')\n",
        "\n",
        "plt.ylabel(\"episode_reward\")\n",
        "plt.xlabel(\"episode (N_steps per episode = 32)\")\n",
        "plt.legend(['DQN','DRQN'])\n",
        "plt.title(\"Unequal match and non-match\")"
      ],
      "metadata": {
        "id": "v1CkmCXg1uLI"
      },
      "execution_count": null,
      "outputs": []
    }
  ],
  "metadata": {
    "colab": {
      "collapsed_sections": [],
      "name": "Copy of Modelling Cognitive Tasks",
      "provenance": [],
      "toc_visible": true
    },
    "kernel": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "accelerator": "GPU",
    "gpuClass": "standard"
  },
  "nbformat": 4,
  "nbformat_minor": 0
}