This repository contains implementations of selected exercises and examples from the "Reinforcement Learning: An Introduction" textbook by Sutton & Barto. Each script or folder corresponds to a different exercise or example, exploring core concepts of reinforcement learning such as value iteration, Monte Carlo methods, Temporal Difference learning, and control problems.
-
4.9_gamblers_problem_value_it.py: Solves the Gambler's Problem using value iteration, exploring optimal policy and state value computation.
-
5.12_racetrack_mc.py: Implements the Racetrack problem using Monte Carlo methods for policy control and improvement.
-
6.9_windy_gridworld_td_sarsa.py: Solves the Windy Gridworld using the SARSA algorithm, a form of Temporal Difference learning, to find an optimal policy. Also solved using Q-learning.
-
7.2_n_step_td_comparison.py: Compares different n-step Temporal Difference (TD) learning methods to study their convergence and performance.
-
ex10.1_mountain_car/: A folder containing scripts and code related to the Mountain Car problem, exploring control using various reinforcement learning algorithms. Includes tiling for extracting features and solved using the semi-gradient SARSA algorithm with linear function approximation.
All output results, including plots and data, are saved in the results folder.
Each script can be run individually to observe the implementation and results of the corresponding RL exercise. For example:
python 4.9_gamblers_problem_value_it.py