murphylmf / breakout-drl

Implemented Naive DQN, Dueling DQN, Priortized Experience Replay for Atari-Breakout

breakout-drl

This repositories contains implementation Naive DQN, Dueling DQN, Prioritized Experience Replay for Atari-Breakout

We obatin our best score of 834 via PER+DDQN+Transfer Learning, and the corresponding weights is stored in best.weights. You can play with it in ./best.ipynb

Getting Started

You can install our environments with

pip install -r requirements.txt

Go into any directory you want to play with(e.g. ./per_ddqn_transfer)

cd per_ddqn_transfer

And run

./run.sh

Evaluating

You can certainly see how our model play the game in best.ipynb.

If you want to resume training, please follow this instruction.

1. Set a variable storing the path to the weights you want to load.(e.g. restore = './best.weights')
2. Set EPS_START and EPS_END to 0 in main.py.
3. Parse the path into $Agent.__init__()$, see utils_drl.py for more details.
4. run the run.sh script.