Regensburg 2025 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
DY: Fachverband Dynamik und Statistische Physik
DY 33: Machine Learning in Dynamics and Statistical Physics I
DY 33.2: Talk
Thursday, March 20, 2025, 09:45–10:00, H47
Finite integration time drives optimal dynamic range into subcritical regime — Sahel Azizpour1,2, Viola Priesemann3,4, •Johannes Zierenberg3,4, and Anna Levina1,2 — 1Eberhard Karls University of Tübingen, Germany — 2Max Planck Institute for Biological Cybernetics, Tübingen, Germany — 3Max Planck Institute for Dynamics and Self Organisation, Göttingen, Germany — 4Institute for the Dynamics of Complex Systems, University of Göttingen, Germany
Sensitivity to small changes in the environment is crucial for many real-world tasks, enabling living and artificial systems to make correct behavioral decisions. It has been shown that such sensitivity is maximized when a system operates near the critical point of a second-order phase transition. However, proximity to criticality introduces large fluctuations and diverging timescales. Hence, it would require impractically long integration periods to leverage the maximal sensitivity. Here, we analytically and computationally demonstrate how the optimal tuning of a recurrent neural network is determined given a finite integration time. Rather than maximizing the theoretically available sensitivity, we find networks to attain different sensitivity depending on the time available. Consequently, the optimal dynamic regime shifts from critical to subcritical when integration times are finite, highlighting the necessity of incorporating finite-time considerations into studies of information processing.
Keywords: Statistical Physics; Criticality; Reservoir Computing; Dynamic Range