Deep Predictor-Corrector Networks for Robust Parameter Estimation in Non-autonomous System with Discontinuous Inputs

TL;DR

DePCoN introduces a multi-scale learning framework that stabilizes optimization in non-autonomous systems with discontinuous inputs, improving robustness and convergence speed over existing methods.

Contribution

The paper presents DePCoN, a novel predictor-corrector network that learns scale-consistent updates, integrating smoothing into the learning process for better stability.

Findings

DePCoN outperforms existing methods on biological and ecological benchmarks.

It achieves faster convergence and reduced hyperparameter sensitivity.

The approach offers a general principle for non-smooth optimization stabilization.

Abstract

Learning under non-smooth objectives remains a fundamental challenge in machine learning, as abrupt changes in conditioning variables can induce highly non-smooth loss landscapes and destabilize optimization. This difficulty is particularly pronounced in non-autonomous dynamical systems driven by discontinuous inputs, where widely used optimization methods, including recent neural smoothing approaches, exhibit unreliable convergence or strong hyperparameter sensitivity. To address this issue, we propose Deep Predictor-Corrector Networks (DePCoN), a multi-scale learning framework that stabilizes optimization by learning scale-consistent parameter update rules across a hierarchy of smoothed inputs. Rather than treating smoothing as a fixed preprocessing choice, DePCoN integrates smoothing into the learning dynamics itself through a learned predictor-corrector mechanism. Across biological and ecological benchmarks with discontinuous inputs, DePCoN consistently achieves more robust and faster convergence than existing methods while substantially reducing sensitivity to hyperparameter choices. Beyond dynamical systems, our approach provides a general learning principle for stabilizing optimization under non-smooth objectives.