This project investigates the intuitions/ideas behind Double DQN, and evaluate how much it can improve Q-value overestimation and agent performance. We aim to describe how the learning/update process in Double DQN ends up with better Q-value estimates and agent performance when comparing to that of DQN.

gym environnement to simulate the energetic behaviour of a real estate

Lunar Lander envitoment of gymnasium solved using Double DQN and D3QN

This repo contains lessons, notes, assignments and a final project from the reinforcement learning lesson at the ozyegin university.

Dueling Network Architecture Implementation for Deep Reinforcement Learning

Using DQN and Double DQN to play Flappy Bird

RL-based agent for playing Atari Pong

Implementation of RL algorithms to beat Atari 2600 games

This project uses deep RL to train an agent that can play Atari game named Space Invaders.

Experiments on Breakout game applying Reinforcement Learning techniques

Deep Reinforcement Learning using pytorch - Bananas

Project1 Navigation

Training reinforcement learning algorithm to play snake.

This project provides a comprehensive understanding of reinforcement learning, focusing on Deep Q-Learning (DQN). It involves exploring the OpenAI Gym library, implementing DQN from DeepMind's seminal paper, and enhancing the DQN algorithm for improved performance and stability.

Implementation of the bit-flipping experiment in the HER paper

Using Double Q Networks with experience replay to solve Cartpole v0 in just 184 episodes, implemented in Tensorflow 2.

Udacity Deep Reinforcement Learning Nanodegree Program - Navigation Control

Solving CartPole using Distributional RL

An intelligent PACMAN agent built using Reinforcement learning.

Build and test DRL algorithms in different environments