Reinforcement Learning for Efficient Identification of Soliton System Parameters Across Expansive Domains

Abstract Optical solitons in mode-locked fiber lasers and optical communication links have various applications. The study of transmission modes of optical solitons necessitates the investigation of the relationship between the equation parameters and soliton evolution employing deep learning techniques. However, the existing identification models exhibit a limited parameter domain search range and are significantly influenced by initialization. Consequently, they often result in divergence toward incorrect parameter values. This study harnessed reinforcement learning to revamp the iterative process of the parameter identification model. By developing a two-stage optimization strategy, the model could conduct an accurate parameter search across arbitrary domains. The investigation involved several experiments on various standard and higher-order equations, illustrating that the innovative model overcame the impact of initialization on the parameter search, and the identified parameters are guided toward their correct values. The enhanced model markedly improves the experimental efficiency and holds significant promise for advancing the research of soliton propagation dynamics and addressing intricate scenarios.