Hindsight Experience Replay - Bit flipping experiment in Tensorflow. Check this blog post of mine explaining HER!
This implementation includes:
- Double DQN with 1 hidden layer of size 256.
- Hindsight experience replay memory with "K-future" strategy.
- A very simple bit-flipping evironment as mentioned in the original paper.
To run this code, adjust the hyperparameters from HER.py and type
$ python HER.py
from bash.
- Plot training curve with respect to rewards.
- Train on bit length higher than 30