Desynchronizing of noisy neuron networks using reinforcement learning

Abstract

Mitigating pathological synchrony of neurons in basal ganglia networks was considered as one of the potential mechanisms of deep brain stimulation (DBS) in treating Parkinson's disease. Motivated by reducing the energy of external stimuli, optimal control strategies are presented to regulate DBS waveform so as to mitigate synchronous oscillations of neural networks with fewer energy expenditure. In this paper, the adaptive optimal control of DBS based on reinforcement learning (RL) is designed to desynchronizing phase models of neural populations in the presence of noise. Numerical simulations show the effectiveness of the proposed method. Moreover, the influence of noise intensity on the control performance of the controller is analyzed.