In-context learning to predict critical transitions in dynamical systems

TL;DR

This paper introduces TipPFN, an in-context learning framework trained on synthetic data to predict critical transitions in dynamical systems, outperforming traditional methods in real-world scenarios.

Contribution

The authors develop a novel in-context learning approach, TipPFN, trained on synthetic bifurcation-based data to detect critical transitions across various systems.

Findings

TipPFN achieves state-of-the-art early detection of critical transitions.

It performs well on unseen tipping regimes and real-world data.

The method is robust across different contexts and complexities.

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.