01. Fundamentals of Reinforcement Learning/1.10. Model-Based and Model-Free Learning .ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model-Based and Model-free learning\n",
    "\n",
    "**Model-based learning** - In model-based learning, an agent will have a complete description of the environment. That is, we learned that the transition probability tells the probability of moving from a state  to the next state  by performing an action  and the reward function tells the reward we would obtain while moving from a state  to the next state  by performing an action . When the agent knows the model dynamics of the environment, that is, when the agent knows the transition probability of the environment then it is called model-based learning. Thus, in model-based learning the agent uses the model dynamics for finding the optimal policy. \n",
    "\n",
    "**Model-free learning** - When the agent does not know the model dynamics of the environment then it is called model-free learning. That is, In model-free learning, an agent tries to find the optimal policy without the model dynamics. \n",
    "\n",
    "Thus, to summarize, in a model-free setting, the agent learns optimal policy without the model dynamics of the environment whereas, in a model-based setting, the agent learns the optimal policy with the model dynamics of the environment."
   ]
  }
 ],
 "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": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}