In-context learning to predict critical transitions in dynamical systems

2026-05-12

AI Summary

This work addresses the challenge of early warning for critical transitions in real-world systems, where such events are rare and heavily obscured by noise, rendering traditional statistical methods and existing deep learning models ineffective. The authors propose TipPFN, a novel framework that introduces in-context learning to critical transition prediction for the first time. By pretraining a Prior-Data Fitted Network (PFN) on diverse synthetic data generated from stochastic dynamical systems grounded in canonical bifurcation scenarios, TipPFN achieves zero-shot generalization to unseen transition mechanisms. The method demonstrates robust and state-of-the-art early warning performance across simulated-to-real transfer settings, unknown transition types, and real observational data, significantly outperforming current approaches.

Abstract

Critical transitions - abrupt, often irreversible changes in system dynamics - arise across human and natural systems, often with catastrophic consequences. Real-world observations of such shifts remain scarce, preventing the development of reliable early warning systems. Conventional statistical and spectral indicators, such as increasing variance, tend to fail under realistic conditions of limited data and correlated noise, whereas existing deep learning classifiers do not extrapolate beyond their training data distribution. In this work, we introduce TipPFN, an in-context learning (ICL) framework that uses a prior-data fitted network to infer a system's proximity to a critical transition. Trained on our novel synthetic data generator, which is based on canonical bifurcation scenarios coupled to diverse, randomized stochastic dynamics, TipPFN flexibly capitalizes on contexts of various sizes, complexity and dimensionalities. We demonstrate robust, state-of-the-art early detection of critical transitions in previously unseen tipping regimes, sim-to-real examples, and real-world observations in both ICL and zero-shot settings.

Problem

• critical transitions
• early warning systems
• in-context learning
• dynamical systems
• limited data

Innovation

• in-context learning
• critical transitions
• TipPFN
• bifurcation dynamics
• early warning systems