Predicting discrete-time bifurcations with deep learning

Many natural and man-made systems are prone to critical transitions—abrupt and potentially devastating changes in dynamics. Deep learning classifiers can provide an early warning signal for critical transitions by learning generic features of bifurcations from large simulated training data sets. This study trains a deep learning classifier to predict the five local discrete-time bifurcations of codimension-one, testing it on simulation data from discrete-time models (physiology, economics, ecology) and experimental data (chick-heart aggregates). The classifier demonstrates higher sensitivity and specificity than traditional early warning signals, accurately predicting bifurcation types, particularly period-doubling, Neimark–Sacker, and fold bifurcations. The research suggests deep learning's potential to revolutionize critical transition monitoring.

Thomas M. Bury, Daniel Dylewsky, Chris T. Bauch, Madhur Anand, Leon Glass, Alvin Shrier, Gil Bub