Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster A in Poster Session A: Tuesday, August 4, 9:30 – 11:15 am, Kimmel Center, Shorin & Rosenthal Rooms

Fast dynamical similarity analysis

Arman Behrad¹, Mitchell Ostrow², Mohammad Taha Fakharian³, Ila R Fiete², Christian Beste⁴, Shervin Safavi¹; ¹Technische Universität Dresden, ²Massachusetts Institute of Technology, ³Okinawa Institute of Science and Technology Graduate University, ⁴TU Dresden

Presenter: Arman Behrad

To understand how nonlinear dynamical systems process information (e.g., artificial neural networks and neural circuits), it is essential to compare the underlying dynamics of diverse systems at scales (e.g., a large pool of neural networks with diverse architectures and large-scale recordings from neural circuits). Existing similarity methods remain inadequate for this purpose: geometric methods are computationally efficient but fail to capture the governing dynamics, whereas dynamical similarity methods are more faithful but often too computationally expensive. Here, we introduce fast Dynamical similarity analysis (fastDSA), a computationally efficient and accurate method for comparing nonlinear dynamical systems. FastDSA combines automatic rank detection, efficient alignment of dynamical flow fields, and Koopman embeddings. Across benchmark nonlinear systems and recurrent network models, it is insensitive to arbitrary coordinate choices but sensitive to real differences in system dynamics, extending the practical scope of dynamical similarity analysis.

Topic Area: Methods, Tools, Theory & Neural Coding