examples/Reinforcement Learning.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Reinforcement learning\n",
    "\n",
    "In this example, we are going to use a evolutionary algorithm to train a dense neural network for the gym's [BipedalWalker environment](https://gym.openai.com/envs/BipedalWalker-v2/)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![](RL-Figure1.jpg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import gym\n",
    "import numpy as np\n",
    "import evolvepy as ep\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Evaluator\n",
    "\n",
    "To speed up the individuals evaluation, we are going to use multiple process. Because of that, we need to define a class that will act on what function we evaluate, which, as Jupyter Notebook doesn't handle multiprocessing well, we need to define it in a separate file.\n",
    "\n",
    "The [bipedal_evaluator.py](bipedal_evaluator.py) file contains:\n",
    "\n",
    "-  BipedalWalkerFitnessFunction: a GymFitnessFunction, evaluates the individual by making him control the agent through the environment, using the total reward obtained as fitness."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from bipedal_evaluator import BipedalWalkerFitnessFunction\n",
    "\n",
    "evaluator = ep.evaluator.ProcessEvaluator(BipedalWalkerFitnessFunction)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As the environment contains random properties, the final score can vary, generating harmful noise in fitness. To deal with this problem, we're going to run 10 assessments on the same individual and then averaging and discarding the highest and lowest."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "multiple_evaluation = ep.evaluator.MultipleEvaluation(evaluator, 10, discard_max=True, discard_min=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Generator\n",
    "\n",
    "Let's start by defining the descriptor of our individuals. Each individual will own the layer weights and bias. The number of neurons is described in the \"sizes\" variable. Let's put as range of each layer (-1, 1)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import nn\n",
    "descriptor = nn.create_descriptor(input_size=24, output_size=4, units=[20,20])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As we are going to evaluate individuals using multiple processes, it is more efficient to use a population size multiple of the number of processors, using all of them at all times:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "import multiprocessing as mp\n",
    "\n",
    "population_size = (100//mp.cpu_count())*mp.cpu_count()\n",
    "population_size"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's define a generator with elitism:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "first = ep.generator.Layer()\n",
    "combine = ep.generator.CombineLayer(ep.generator.selection.tournament, ep.generator.crossover.one_point)\n",
    "mutation = ep.generator.mutation.NumericMutationLayer(ep.generator.mutation.sum_mutation, 1.0, 0.9, (-0.5, 0.5))\n",
    "filter0 = ep.generator.FilterFirsts(int(np.floor(0.95*population_size)))\n",
    "sort = ep.generator.Sort()\n",
    "filter1 = ep.generator.FilterFirsts(int(np.ceil(0.05*population_size)))\n",
    "concat = ep.generator.Concatenate()\n",
    "\n",
    "first.next = combine\n",
    "combine.next = mutation\n",
    "mutation.next = filter0\n",
    "filter0.next = concat\n",
    "\n",
    "first.next = sort\n",
    "sort.next = filter1\n",
    "filter1.next = concat\n",
    "\n",
    "generator = ep.generator.Generator(first_layer=first, last_layer=concat, descriptor=descriptor)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To avoid getting stuck in some generation, we're also going to use dynamic mutation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "dyn_mut = ep.callbacks.DynamicMutation([mutation.name], refinement_patience=5, exploration_patience=5, refinement_steps=5)\n",
    "\n",
    "evolver = ep.Evolver(generator, multiple_evaluation, population_size, [dyn_mut])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Evolve and results\n",
    "\n",
    "Let's evolve our generation:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "hist, last_pop = evolver.evolve(500)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(hist.max(axis=1))\n",
    "plt.plot(np.mean(hist, axis=1))\n",
    "plt.xlabel(\"Generation\")\n",
    "plt.ylabel(\"Fitness\")\n",
    "plt.title(\"Evolution History\")\n",
    "plt.legend([\"Best\", \"Mean\"])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Looking at the fitness history, we can see that the algorithm managed to evolve our population. Realize how fitness has a noise, not using multiple assessments makes it even greater, hindering evolution. Variation in average fitness is one of the side effects of dynamic mutation.\n",
    "\n",
    "![](RL-Figure2.jpg)\n",
    "\n",
    "It is interesting to note that our population changes drastically at times. This phenomenon is called evolutionary steps.\n",
    "\n",
    "We were able to see the best individual reached a maximum fitness of ~70. The score needed to go through the environment is 300. We could further improve our agent by:\n",
    "- Evolve for more generations\n",
    "- Changing the Neural Network Architecture\n",
    "- Change the generation or evaluation scheme"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can record a video of our best individual:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "best = last_pop[np.argmax(hist[-1])]\n",
    "test_evaluator = BipedalWalkerFitnessFunction(save=True)\n",
    "test_evaluator([best])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Clicking on the picture below you can see a video with the best individuals after each evolutionary step:\n",
    "\n",
    "[![](RL-Figure1.Jpg)](https://youtu.be/WwIoEMIe3ic)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finally, by adding a WandbCallback to the evolver, we can generate a more detailed log of the evolution: [example log](https://wandb.ai/eltoncn/EvolvePy%20Example/runs/2o9vemwl?workspace=user-eltoncn)."
   ]
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "2db524e06e9f5f4ffedc911c917cb75e12dbc923643829bf417064a77eb14d37"
  },
  "kernelspec": {
   "display_name": "Python 3.8.5 64-bit",
   "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": "3.8.5"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}