Learning of Switched Nonlinear Dynamical Systems

#cyber-physical-systems #nonlinear-dynamical-systems

Cyber-physical systems can be found in various applications and identifying their hybrid dynamics from data is a challenging task due to the complexity of capturing both continuous and discrete components. Most hybrid system identification methods rely on passive learning, thus considering a fixed dataset without interacting with the system-under-learning. Active learning of hybrid systems remain less explored. In this talk we present  a  learning approach for nonlinear switched systems where switchings are state-dependent. This approach combines passive and active learning. The passive method involves solving an optimisation problem over a dataset, while the active method incrementally learns the hybrid dynamics by leveraging equivalence queries to locate discrepancies between the learned model and the true system and by exploiting counterexamples to refine the model. 

• Add Event to Calendar
  iCal
  Google Calendar

• Aalborg University
• Fredrik Bajers Vej 7
• Aalborg, Nordjyllands Amt
• Denmark

---

• Contact Event Host

• Starts 31 July 2025 10:00 PM UTC
• Ends 20 August 2025 10:00 PM UTC
• No Admission Charge

•