Initial results of Vanilla DQN trained with the pong environment.
Implemenentation of DQN training using PyTorch (uses a bit of starter code from https://github.com/berkeleydeeprlcourse/homework)
Exploration of various algorithms and implementations starting from what was presented in https://www.cs.toronto.edu/~vmnih/docs/dqn.pdf.
Additionally exploring the impact of learning with:
- Vanilla DQN (https://www.cs.toronto.edu/~vmnih/docs/dqn.pdf) (working)
- Multiple parallel agents (https://arxiv.org/pdf/1602.01783.pdf) (testing)
- Double DQN (https://arxiv.org/pdf/1509.06461.pdf) (testing)
- TBD.