pacman-env.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "# `MsPacman-v0` environment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "import gym"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[2017-10-18 21:17:43,239] Making new env: MsPacman-v0\n"
     ]
    }
   ],
   "source": [
    "env = gym.make('MsPacman-v0')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "state = env.reset()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(210, 160, 3)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "state.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "You will notice that `env.reset()` returns a large array of numbers. To be specific, you can enter `state.shape` to show that our current state is represented by a `210x160x3` Tensor. This represents the `height`, `length`, and the three `RGB` color channels of the Atari game or, simply put, the pixels. \n",
    "\n",
    "Let's visualize the environment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "# env.render()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "### Possible actions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "9"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "env.action_space.n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "This will show that we have nine possible actions: integers 0-8. It’s important to remember that an agent should have no idea what these actions mean; its job is to learn which actions will optimize reward. But, for our sake, we can:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(0, 'NOOP'),\n",
       " (1, 'UP'),\n",
       " (2, 'RIGHT'),\n",
       " (3, 'LEFT'),\n",
       " (4, 'DOWN'),\n",
       " (5, 'UPRIGHT'),\n",
       " (6, 'UPLEFT'),\n",
       " (7, 'DOWNRIGHT'),\n",
       " (8, 'DOWNLEFT')]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(enumerate(env.env.get_action_meanings()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "### Random solution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "state = env.reset()\n",
    "reward, done, info = None, None, None\n",
    "while not done:\n",
    "    state, reward, done, info = env.step(env.action_space.sample())\n",
    "    #env.render()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "This completely random policy will get a few hundred points, at best, and will never solve the first level.\n",
    "\n",
    "Continuing on, we cannot use our basic Q table algorithm because there is a total of 33,600 pixels with three RGB values that can have a range from 0 to 255. It’s easy to see that things are getting extremely complicated; this is where deep learning comes to the rescue. Using techniques such as convolutional neural networks or a DQN, a machine learning library is able to take the complex high-dimensional array of pixels, make an abstract representation, and translate that representation into a optimal action."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "## Conclusion"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "The field of reinforcement learning is rapidly expanding with new and better methods for solving environments—at this time, the A3C method is one of the most popular. Reinforcement learning will more than likely play an important role in the future of AI and continues to produce very interesting results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}