Repository for work related to Udacity's Deep Reinforcment Learning Nanodegree. See individual project subfolders for more detailed descriptions.
Train an agent to gather bananas in the Unity Banana Collector environment. Demonstrates the Deep Q-Learning algorithm and the effects of various hyperparameter settings. Note that this version of the agent does not use image information from the environment, but rather uses state information directly in a 37-dimensional vector that includes ego-information and ray-based information about its surroundings.
Train a robotic arm to follow a moving target region in space. Demonstrates the Deep Deterministic Policy Gradient Algorithm - an actor-critic method for continuous action spaces.
Train two agents to play a game of tennis in the Unity Tennis environment. Demonstrates DDPG with prioritized experience replay with multiple agents.