Real-Time Recurrent Reinforcement Learning

arXiv:2311.04830v2 Announce Type: replace
Abstract: In this paper we propose real-time recurrent reinforcement learning (RTRRL), a biologically plausible approach to solving discrete and continuous control tasks in partially-observable markov decision processes (POMDPs). RTRRL consists of three parts: (1) a Meta-RL RNN architecture, implementing on its own an actor-critic algorithm; (2) an outer reinforcement learning algorithm, exploiting temporal difference learning and dutch eligibility traces to train the Meta-RL network; and (3) random-feedback local-online (RFLO) learning, an online automatic differentiation algorithm for …

abstract actor actor-critic algorithm architecture arxiv continuous control cs.lg cs.ne cs.sy decision eess.sy markov meta observable paper processes real-time reinforcement reinforcement learning rnn tasks type