Towards neural reinforcement learning for large deviations in non-equilibrium systems with memory

We introduce a reinforcement learning method for a class of non-Markov systems; our approach extends the actor–critic framework given by Rose et al (2021 New J. Phys. 23 013013) for obtaining scaled cumulant generating functions characterizing the fluctuations. The actor–critic is implemented using neural networks; a particular innovation in our method is the use of an additional neural policy for processing memory variables. We demonstrate results for current fluctuations in various memory-dependent models with special focus on semi-Markov systems where the dynamics is controlled by nonexponential interevent waiting time distributions.