Disertation project at the University of Sheffield.
The project focuses on the development of conventional and novel deep reinforcement learning agents, including agents employing direct feedback alignment for training the deep neural network used in predicting the action-value function.
The codes for this project are split into three classes:
- agents for classical reinforcement learning tasks
- agents for classical control problems
- agents for Atari games
For more information on the project, access the text of my dissertation here: (link TBA)
For information about the various projects, read the README in each project.