This repository contains a set of algorithms and gyms for educational purposes and reinforcement learning experiments. The primary goal is to implement everything from scratch to gain a deeper understanding of RL.
Algorithms:
- DQN
- DoubleDQN
- DuelingDQN
GYMs:
- Snake
TODO:
- RainbowDQN implementation
- Platformer environment
- PPO (Proximal Policy Optimization) implementation
Use python 3.10 and following packages:
$ pip install -r requirements.txt
Run script:
$ python run.py --param_file YOUR_PARAM_FILE --mode train --plot_results --display
For example
python run.py --param_file ./configs/double_dqn_snake.yaml --mode train --plot_results --display
Parameters_search:
python run.py --param_file ./configs/double_dqn_snake.yaml --mode parameters_search
Test:
python run.py --param_file ./configs/double_dqn_snake.yaml --mode test --display
DQN Result
Double DQN Result
Dueling DQN Result
