2501.10339

Total: 1

#1 Principled model selection for stochastic dynamics [PDF] [Copy] [Kimi] [REL]

Authors: Andonis Gerardos, Pierre Ronceray

Complex dynamical systems, from macromolecules to ecosystems, are often modeled by stochastic differential equations (SDEs). To learn such models from data, a common approach involves decomposing the SDE into a linear combination of basis functions. However, this can induce overfitting due to the proliferation of parameters. To address this, we introduce Parsimonious Stochastic Inference (PASTIS), a principled method that removes superfluous parameters from SDE models by combining likelihood-estimation statistics with extreme value theory. We benchmark it against existing methods and show that it reliably selects the exact minimal models from large libraries of functions, even with a low sampling rate or measurement error. We show that it extends to stochastic partial differential equations and demonstrate applications to the inference of ecological networks and reaction-diffusion dynamics.

Subjects: Soft Condensed Matter , Statistical Mechanics , Data Analysis, Statistics and Probability , Machine Learning

Publish: 2025-01-17 18:23:16 UTC