Q-learning based asynchronous Boolean control networks stabilization with data loss.

Abstract

In real-world networks, data loss and asynchronous updates are often unavoidable. To this end, this paper combines Q-learning and flip control strategies to investigate the stabilization of asynchronous Boolean networks (ABNs) with data loss. Firstly, using the semi-tensor product (STP) tool, an algebraic formulation of the model is derived. To facilitate subsequent theoretical analysis, the original system is constructed as a corresponding augmented system and their equivalence is proved. Based on this, the stabilization criterion of the system under flip control is established. Finally, Q-learning is introduced to design flip sequences for achieving the stabilization goal and the effectiveness of proposed methods is verified via two biological examples.