Re-purposing Compact Neuronal Circuit Policies to Govern Reinforcement Learning Tasks
(AAAI 2019 submission)
Arxiv Link: https://arxiv.org/abs/1809.04423
“Inspired by the structure of the nervous system of the soil-worm, C. elegans, we introduce Neuronal Circuit Policies (NCPs) as a novel, a recurrent neural network instance with liquid time-constants, universal approximation capabilities and interpretable dynamics.”
This paper theoretically shows that neuronal circuit policies (an RNN inspired by the nervous system of a soil-worm) perform as good as deep neural network policies with the advantage of realizing interpretable dynamics (basically means internal workings of the system).
They have open-sourced their code: