Aperiodic Sampled-Data Stabilization of Probabilistic Boolean Control Networks: Deep Q-learning Approach with Relaxed Bellman Operator

Successful control of gene regulatory networks (GRNs) can help obtain potential gene treatments by making therapeutic interventions. In this work, GRNs are modeled as probabilistic Boolean control networks (PBCNs) and strategies to devise an optimal feedback control are discussed. Model-free reinforcement learning (RL) based control is proposed in order to minimize model design efforts and regulate GRNs with high complexities. We make use of a deep Q-learning protocol to stabilize PBCNs in an aperiodic control framework. Finally, main results are validated using computer simulations.