One-shot learning for the complex dynamical behaviors of weakly nonlinear forced oscillators

Teng Ma, Luca Rosafalco, Wei Cui, Lin Zhao, Attilio Frangi|April 16, 2026arXiv

Key Takeaway

You can train a model on a single time-series measurement to predict the full range of behaviors in nonlinear oscillating systems, dramatically reducing the data needed to characterize real-world devices like MEMS sensors.

Summary

This paper presents MEv-SINDy, a machine learning method that learns the governing equations of complex nonlinear systems from just a single measurement.

training reasoning

Key Terms

sparse-autoencoder one-shot-learning harmonic-reasoning extrapolative-prediction