examples/Basic tutorial on Boolearn Reasoning problem.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python2.7/dist-packages/matplotlib/font_manager.py:273: UserWarning: Matplotlib is building the font cache using fc-list. This may take a moment.\n",
      "  warnings.warn('Matplotlib is building the font cache using fc-list. This may take a moment.')\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "env: THEANO_FLAGS='device=cpu'\n",
      "Couldn't import dot_parser, loading of dot files will not be possible.\n"
     ]
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "import numpy as np\n",
    "from matplotlib import pyplot as plt\n",
    "\n",
    "\n",
    "#theano imports. Setting up to use CPU\n",
    "%env THEANO_FLAGS='device=cpu'\n",
    "import theano\n",
    "import lasagne"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# This tutorial explains the basic pipeline of Agentnet experiment\n",
    "* experiment setup\n",
    "* designing agent\n",
    "* interacting with environment\n",
    "* computing losses\n",
    "* training"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### A friendly note\n",
    "This example does not use external environments. If you want to dive into openai gym, you may as well [start here](http://bit.ly/29EvHvZ)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Experiment setup\n",
    "\n",
    "Load a simple experiment environment and print its description"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Boolean reasoning is a dummy experiment setup that requires agent to make advantage of\n",
      "a simple logical formula in order to maximize expected reward.\n",
      "\n",
      "The world agent exists in has a number of boolean hidden factors:\n",
      "X1, X2, X3, Y1, Y2.\n",
      "\n",
      "The factors are not independent. Namely,\n",
      " - Y1 is true if X1\n",
      " - Y2 is true if not X1\n",
      "\n",
      "In the initial moment of time, agent knows nothing about any of them.\n",
      "At each turn, agent may decide to\n",
      " - \"open\" one of the hidden factors.\n",
      "   - if the factor turns out to be 1, agent receives +1 reward for X*, +3 for Y*\n",
      "   - Otherwise, the reward equals -1 for X*, -3 for Y*\n",
      "   - checking a single factor more than once a session will result in -0.5 reward for every attempt but for first one\n",
      " - decide to quit session\n",
      "   - yields reward of 0 and ends the interaction.\n",
      "   - all farther actions will have no effect until next session\n",
      "\n",
      "It is expected, that in order to maximize it's expected reward, the agent\n",
      "will learn the policy\n",
      "1. Check X1\n",
      "2. If X1, check Y1, else Y2\n",
      "\n",
      "The experiment setup contains a single class BooleanReasoningEnvironment that\n",
      "implements both BaseEnvironment and BaseObjective.\n",
      "\n"
     ]
    }
   ],
   "source": [
    "import agentnet.experiments.boolean_reasoning as experiment\n",
    "# instantiate an experiment environment with default parameters\n",
    "env = experiment.BooleanReasoningEnvironment()\n",
    "\n",
    "print(experiment.__doc__)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Agent setup\n",
    "* An agent implementation has to contain three parts:\n",
    "  * Memory layer(s)\n",
    "    * in this case, a single one-step RNN\n",
    "  * Q-values evaluation layers\n",
    "    * in this case, a lasagne dense layer based on memory layer\n",
    "  * Resolver - an action picker layer\n",
    "    * in this case, the resolver has epsilon-greedy policy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from agentnet.resolver import EpsilonGreedyResolver\n",
    "from agentnet.memory.rnn import RNNCell\n",
    "from agentnet.agent import Agent"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# hidden neurons\n",
    "n_hidden_neurons = 64\n",
    "\n",
    "observation_size = (None,) + tuple(env.observation_shapes)\n",
    "\n",
    "observation_layer = lasagne.layers.InputLayer(observation_size, name=\"observation_input\")\n",
    "prev_state_layer = lasagne.layers.InputLayer([None, n_hidden_neurons], name=\"prev_state_input\")\n",
    "\n",
    "# memory layer (this isn't the same as lasagne recurrent units)\n",
    "rnn = RNNCell(prev_state_layer, observation_layer, name=\"rnn0\")\n",
    "\n",
    "# q_values (estimated using very simple neural network)\n",
    "q_values = lasagne.layers.DenseLayer(rnn,\n",
    "                                     num_units=env.n_actions,\n",
    "                                     nonlinearity=lasagne.nonlinearities.linear,\n",
    "                                     name=\"QEvaluator\")\n",
    "\n",
    "# resolver uses epsilon - parameter which defines a probability of randomly taken action.\n",
    "epsilon = theano.shared(np.float32(0.1), name=\"e-greedy.epsilon\")\n",
    "resolver = EpsilonGreedyResolver(q_values, epsilon=epsilon, name=\"resolver\")\n",
    "\n",
    "\n",
    "# packing this into agent\n",
    "agent = Agent(observation_layer,\n",
    "              agent_states={rnn:prev_state_layer},\n",
    "              policy_estimators=q_values, \n",
    "              action_layers=resolver)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[rnn0.hid_to_hid.W,\n",
       " rnn0.hid_to_hid.b,\n",
       " rnn0.input0_to_hid.W,\n",
       " QEvaluator.W,\n",
       " QEvaluator.b]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Since it's a lasagne network, one can get it's weights, output, etc\n",
    "weights = lasagne.layers.get_all_params(resolver,trainable=True)\n",
    "weights"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Agent setup in detail\n",
    "* __Memory layers__\n",
    " * One-step recurrent layer\n",
    "     * takes input and one's previous state\n",
    "     * returns new memory state\n",
    "   * Can be arbitrary lasagne layer\n",
    "   * Several one-step recurrent units are implemented in __agentnet.memory__\n",
    "   * Note that lasagne's default recurrent networks roll for several steps at once\n",
    "     * in other words, __using lasagne recurrent units as memory means recurrence inside recurrence__\n",
    " * Using more than one memory layer is explained in farther tutorials\n",
    "\n",
    "\n",
    "* __Q-values evaluation layer__\n",
    " * Can be arbitrary lasagne network\n",
    " * returns predicted Q-values for each action\n",
    " * Usually depends on memory as an input\n",
    "\n",
    "\n",
    "* __Resolver__ - action picker\n",
    " * Decides on what action is taken\n",
    " * Normally takes Q-values as input\n",
    " * Currently all experiments require integer output\n",
    " * Several resolver layers are implemented in __agentnet.resolver__"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Interacting with environment\n",
    "* an agent has a method that produces __symbolic__ environment interaction sessions\n",
    "* such sessions are represented as tensors with dimensions matching pattern [batch_session_i, time_tick, ...]\n",
    "* interactions result in sequences of observations, actions, q-values,etc\n",
    "* one has to pre-define maximum session length.\n",
    "  * environment implements an indicator of whether session has ended by current tick\n",
    "* since this environment also implements Objective methods, it can evaluate rewards for each [batch, time_tick]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# produce interaction sequences of length <= 10\n",
    "(state_seq,), observation_seq, agent_state, action_seq, qvalues_seq = agent.get_sessions(\n",
    "    env,\n",
    "    session_length=10,\n",
    "    batch_size=env.batch_size,\n",
    ")\n",
    "\n",
    "hidden_seq = agent_state[rnn]\n",
    "\n",
    "# get rewards for all actions\n",
    "rewards_seq = env.get_reward_sequences(state_seq, action_seq)\n",
    "\n",
    "# get indicator whether session is still active\n",
    "is_alive_seq = env.get_whether_alive(observation_seq)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Defining loss function\n",
    "\n",
    "Loss consists of two terms\n",
    "- training error on played games\n",
    "    - in our case, this loss is provided by Q-learning\n",
    "- regularization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# get reference Qvalues according to Qlearning algorithm\n",
    "from agentnet.learning import qlearning\n",
    "\n",
    "# gamma - delayed reward coefficient - what fraction of reward is retained if it is obtained one tick later\n",
    "gamma = theano.shared(np.float32(1), name='q_learning_gamma')\n",
    "\n",
    "squarred_Qerror = qlearning.get_elementwise_objective(\n",
    "    qvalues_seq,\n",
    "    action_seq,\n",
    "    rewards_seq,\n",
    "    is_alive_seq,\n",
    "    gamma_or_gammas=gamma)\n",
    "\n",
    "# take sum over steps, average over sessions\n",
    "mse_Qloss = squarred_Qerror.sum(axis=1).mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# impose l2 regularization on network weights\n",
    "from lasagne.regularization import regularize_network_params, l2\n",
    "reg_l2 = regularize_network_params(resolver, l2) * 10**-3\n",
    "\n",
    "loss = mse_Qloss + reg_l2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Compile train and evaluation functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# compute weight updates\n",
    "updates = lasagne.updates.adadelta(loss, weights, learning_rate=0.1)\n",
    "# take sum over steps, average over sessions\n",
    "mean_session_reward = rewards_seq.sum(axis=1).mean()\n",
    "\n",
    "train_fun = theano.function([], [loss, mean_session_reward], updates=updates)\n",
    "\n",
    "compute_mean_session_reward = theano.function([], mean_session_reward)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training loop"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from agentnet.display import Metrics\n",
    "score_log = Metrics()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500 2600 2700 2800 2900 3000 3100 3200 3300 3400 3500 3600 3700 3800 3900 4000 4100 4200 4300 4400 4500 4600 4700 4800 4900\n"
     ]
    }
   ],
   "source": [
    "batch_size = 10\n",
    "\n",
    "for epoch in range(5000):        \n",
    "    \n",
    "    # update resolver's epsilon (chance of random action instead of optimal one)\n",
    "    # epsilon decreases over time\n",
    "    current_epsilon = 0.05 + 0.95 * np.exp(-epoch / 2500.)\n",
    "    resolver.epsilon.set_value(np.float32(current_epsilon))\n",
    "    \n",
    "    # train\n",
    "    env.generate_new_data_batch(batch_size)\n",
    "    loss, avg_reward = train_fun()\n",
    "        \n",
    "    # show current learning progress\n",
    "    if epoch % 100 == 0:\n",
    "        print epoch,\n",
    "        \n",
    "        # estimate reward for epsilon-greedy strategy\n",
    "        avg_reward_current = compute_mean_session_reward()\n",
    "        score_log[\"expected epsilon-greedy reward\"][epoch] = avg_reward_current\n",
    "        \n",
    "        # estimating the reward under assumption of greedy strategy\n",
    "        resolver.epsilon.set_value(0)\n",
    "        avg_reward_greedy = compute_mean_session_reward()\n",
    "        score_log[\"expected greedy reward\"][epoch] = avg_reward_greedy"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluate results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAEKCAYAAAARnO4WAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nvkQikdgJ0uFLJBKJnfD/fcS9loBbka8AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x113a7a450>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "score_log.plot(\"final\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Conclusion\n",
    "\n",
    "We've trained out first agent which plays a fairly trivial game and learned a bit about basic agent structure. \n",
    "\n",
    "To dive deeper into AgentNet, please follow other tutorials."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}