Berlin 2024 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
DY: Fachverband Dynamik und Statistische Physik
DY 42: Focus Session: Computing with Dynamical Systems: New Perspectives on Reservoirs and Applications I – Fundamentals
DY 42.9: Talk
Thursday, March 21, 2024, 12:00–12:15, BH-N 243
Contextual Alignment for Robust Learning in Dynamical Systems — •Max Weinmann1,2 and Miriam Klopotek1 — 1University of Stuttgart, Stuttgart Center for Simulation Science, SimTech Cluster of Excellence EXC 2075, Stuttgart, Germany — 2University of Stuttgart, Interchange Forum for Reflecting on Intelligent Systems, IRIS3D, Stuttgart, Germany
Dynamical systems have exhibited remarkable computational prowess, particularly in prediction tasks like reservoir computing [1]. Achieving efficient computational offloading to the reservoir demands precise alignment of system dynamics with computational requisites. This entails parameter tuning and input transformation to facilitate a natural representation within the system. While the significance of dynamical diversity and system memory is evident, tailoring systems for specific tasks remains challenging [2]; context plays a key role. Our study investigates cellular automaton and other complex model systems. We explore probabilistic and deterministic perturbations to alter system dynamics, offering contextual cues that optimize task-specific information processing capabilities. Our focus extends to identifying fundamental properties enabling a predictable, adaptable alignment, akin to a learning process, fostering robust model creation.
[1] Maass, W., Natschläger, T., and Markram, H. (2002). Real-time computing without stable states: A new framework for neural computation based on perturbations. Neural computation, 14(11), 2531-2560.
[2] Stepney, S. (2012). Nonclassical Computation-A Dynamical Systems Perspective. Handbook of natural computing, 2.
Keywords: Dynamical systems; Reservoir computing; Machine learning; Non-equilibrium; Cellular Automata