Use DQN to solve OpenAI Classic Control environments: MountainCar, Pendulum, CartPole, Acrobot and LunarLander

Code: DQN Classic Control

Reproduction of DeepMind pivotal paper "Playing Atari with Deep Reinforcement Learning" (2013).
Code not tidied, results coming soon.

Code: DQN Atari 2013

Implementation of Simple Bandit Algorithm along with reimplementation of figures 2.1 and 2.2 from the book

Code: Simple Bandit

Implementation of Tracking Bandit Algorithm and recreation of figure 2.3 from the book

Code: Tracking Bandit

Implementation of UCB Bandit Algorithm and recreation of figure 2.4 from the book

Code: UCB Bandit

Implementation of Gradient Bandit Algorithm and recreation of figure 2.5 from the book

Code: Gradient Bandit

Parameter study of bandit algorithms and recreation of figure 2.6 from the book

Code: Summary

Implementation of Iterative policy Evaluation algorithm and demonstration on FrozenLake-v0 environment

Code: Iterative Policy Evaluation

Implementation of Policy Iteration algorithm and demonstration on FrozenLake-v0 environment

Code: Policy Iteration

Implementation of Value Iteration algorithm and demonstration on FrozenLake-v0 environment

Code: Value Iteration

Implementation of First-Visit MC Prediction algorithm, recreation of figure 5.1 and demonstration on Blackjack-v0 environment

Code: First Visit MC Prediction

Implementation of Monte Carlo ES Control algorithm, recreation of figure 5.2 and demonstration on Blackjack-v0 environment

Code: Monte Carlo ES Control

Implementation of On-Policy First-Visit MC Control algorithm and demonstration on Blackjack-v0 environment

Code: On-Policy First-Visit MC Control

Implementation of TD Prediction algorithm, recreation of figure from example 6.2 and demonstration on Blackjack-v0 environment

Code: TD Prediction

Implementation of SARSA algorithm, recreation of figure from example 6.5 and demonstration on Windy Gridworld environment

Code: SARSA

Implementation of Q-Learning algorithm and demonstration on Cliff Walking environment

Code: Q-Learning

Implementation of Gradient MC algorithm, recreation of figure 9.1 and example 9.1 and demonstration on Corridor environment

Code: Gradient Monte Carlo

Implementation of Semi-Gradient TD algorithm, recreation of figure 9.2 and example 9.2 and demonstration on Corridor environment

Code: Semi-Gradient TD

Implementation of Linear Models with Polynomial and Fourier bases, recreation of figure 9.5 and demonstration on Corridor environment

Code: Polynomial and Fourier Bases

Implementation of Linear Model with Tile Coding, recreation of figure 9.10 and demonstration on Corridor environment

Code: Tile Coding

Implementation of Gradient MC with Artificial Neural Network function approximation and demonstration on Corridor environment

Code: Gradient MC ANN

Implementation of Episodic Semi-Gradient SARSA algorithm, recreation of figures 10.1 and 10.2 and demonstration on MountainCar-v0 environment

Code: Episodic Semi-Gradient SARSA

Implementation of REINFORCE algorithm, recreation of figure 13.1 and demonstration on Corridor with switched actions environment

Code: REINFORCE

Implementation of REINFORCE with Baseline algorithm, recreation of figure 13.4 and demonstration on Corridor with switched actions environment

Code: REINFORCE with Baseline

Implementation of One-Step Actor-Critic algorithm, we revisit Cliff Walking environment and show that Actor-Critic can learn the optimal path

Code: One-Step Actor-Critic

Implementation of Policy Parametrization for Continuous Actions with examples on Continuous Bandit

Code: Continuous Actions

Dynamic Programming - Iterative Policy Evaluation, Policy Iteration, Value Iteration

Code: Dynamic Programming

Monte Carlo and Temporal Difference Prediction

Code: MC and TD Prediction

Forward View TD(λ) and Backward View TD(λ) with Eligibility Traces

Code: N-Step and TD(λ) Prediction

On-Policy Control algorithms: MC Control, SARSA, N-Step SARSA, Forward View SARSA(λ), Backward View SARSA(λ) with Eligibility Traces

Code: On-Policy Control

Off-Policy Control algorithms (Expectation Based): Q-Learning, Expected SARSA, Tree Backup Algorithm

Code: Off-Policy Control - Expectation Based

Off-Policy Control algorithms (Importance Sampling) - Imp.Sampl. SARSA, N-Step Imp.Sampl. SARSA, Off-Policy MC Control

Code: Off-Policy Ctrl. - Importance Sampling