Biologically Plausible Variational Policy Gradient with Spiking Recurrent Winner-Take-All Networks
This work addresses the need for task-specific expert knowledge in biologically plausible reinforcement learning, offering a more automated approach for neuromorphic hardware applications.
The paper tackles the problem of heuristic design in reward-modulated spike-timing-dependent plasticity (R-STDP) methods for reinforcement learning by proposing spiking variational policy gradient (SVPG), which derives local learning rules from global policy gradients. In experiments on MNIST classification and Gym InvertedPendulum, SVPG achieves good training performance and better robustness to noise compared to conventional methods.
One stream of reinforcement learning research is exploring biologically plausible models and algorithms to simulate biological intelligence and fit neuromorphic hardware. Among them, reward-modulated spike-timing-dependent plasticity (R-STDP) is a recent branch with good potential in energy efficiency. However, current R-STDP methods rely on heuristic designs of local learning rules, thus requiring task-specific expert knowledge. In this paper, we consider a spiking recurrent winner-take-all network, and propose a new R-STDP method, spiking variational policy gradient (SVPG), whose local learning rules are derived from the global policy gradient and thus eliminate the need for heuristic designs. In experiments of MNIST classification and Gym InvertedPendulum, our SVPG achieves good training performance, and also presents better robustness to various kinds of noises than conventional methods.