hany606 / PMLDL-Project

This repository is for Practical Machine Learning and Deep Learning course project at Innopolis University Fall 2021

PMLDL-Project

This repository is for Practical Machine Learning and Deep Learning course project at Innopolis University Fall 2021

Project directory structure

├── media: photos and gifs

├── algorithms: different algorithms

└── zoo: trained agents

How to Run it?

In order to train

python3 train.py --algo <algo> --env <env>

• replace <algo> with (ppo, sac, or dqn)

• replace <env> with (cartpole, mountaincar, acrobot)

In order to evaluate and render the episode of the agent with the trained model. The default directory is good_zoo directory, to change it you need to change the code. Later, we will make it through cli

python3 run_agent.py --algo <algo> --env <env>

• replace <algo> with (ppo, sac, or dqn)

• replace <env> with (cartpole, mountaincar, acrobot)

A demo for DQN:

Trained Agent for MountainCar:

A demo for PPO:

Trained Agent for cartpole:

Rewards:

Training reward:

Testing reward:

TODO:

• Implement DQN
• Change the implementation of DQN to be more generalized for any environment
• Add support for wandb or Tensorboard
• Create cli interface for run_agent.py
• Implement PPO
• Implement SAC
• Create Benchmark using different envs and between different agorithms