10707-deep-learning-project
Deep Learning Course Project
| DQN 1 vs 1 | DQN 1 vs 2 | DQN 2 vs 1 |
|---|---|---|
 |
 |
 |
This repository contains our implementation of DQN, DDPG, and MADDPG that works on a slightly modified version of the predator-pray environment. It also contains our results, including trained weights and training rewards and losses.
Here is a summary of the folders:
multiagent-particle-envs/: This folder contains our implementation of DQN, DDPG, and MADDPG agents (dqn.py,ddpg.py,maddpg.py). It also contains the main file used for training and testing the agents (dqn_tag.py,ddpg_tag.py,maddpg_tag.py).multiagent-particle-envs/multiagent/scenarios/: This folder contains the modified version of the predator-pray scenario (simple_tag_guided_1v1.py,simple_tag_guided_1v2.py,simple_tag_guided_2v1.py).report/: This folder contains the latex files for our final report.results/: This folder contains the trained weights of our DQN, DDPG, and MADDPG models, and the statistics collected during training.