Reinforcement learning (RL) tasks are challenging to implement, execute and test due to algorithmic instability, hyper-parameter sensitivity, and heterogeneous distributed communication patterns. We argue for the separation of logical component composition, backend graph definition, and distributed execution. To this end, we introduce RLgraph, a library for designing and executing reinforcement learning tasks in both static graph and define-by-run paradigms. The resulting implementations are robust, incrementally testable, and yield high performance across different deep learning frameworks and distributed backends.
@article{arxiv.1810.09028,
title = {RLgraph: Modular Computation Graphs for Deep Reinforcement Learning},
author = {Michael Schaarschmidt and Sven Mika and Kai Fricke and Eiko Yoneki},
journal= {arXiv preprint arXiv:1810.09028},
year = {2019}
}