中文无码

报告人：李彪 教授（宁波大学）

时间：2026年01月27日 15:00-

腾讯会议ID：495 681 894   密码：5258

摘要：In this paper, we propose nonlinear interaction multi-domain physics-informed neural networks (NI-MPINNs) by combining the nonlinear interaction and domain decomposition techniques. This model is employed to investigate the dynamic behavior of hybrid rogue wave and breather solutions, as well as to facilitate parameter discovery within the context of the nonlinear Schrödinger equation (NLSE). The NI-MPINNs model is employed to learn various hybrid wave solutions, including the hybrid first-order rogue wave and first-order breather solution, hybrid second-order rogue wave and first-order breather solution, as well as the hybrid first-order rogue wave and second-order breather solution. The experimental results display that, in comparison to classical PINNs, the NI-MPINNs model enhances prediction accuracy by three orders of magnitude. For inverse problems, the NI-MPINNs model effectively identifies unknown parameters in the NLSE under both noisy and noise-free conditions, addressing the limitations of classical PINNs in parameter identification for the NLSE and demonstrating strong robustness.

邀请人：刘磊

欢迎广大师生积极参与！