Meta learner A3C algorithm with a neural relational memory module. It's currently solving a very simple contextual bandits problem. Further experimentation with some other problems are expected in a near future.
To install the dependencies run:
pip install -r requeriments.txt
To run the algorithm:
python agent.py
[1] Original A3C algorithm taken from https://github.com/awjuliani Arthur Juliani repo.
[2] "Relational Recurrent Neural Networks", Santoro et al., 2018.