Efficient method detects sharp regime changes in simulation models
Efficient regime boundary detection in discontinuous simulation models
Computational Engineering, Finance, and Science
Summary
Simulations often change behavior suddenly when input values shift, making it hard to predict overall patterns. The authors show that treating these changes as a yes-or-no classification problem helps find where the changes happen more efficiently. They created a way to pick new test points smartly around the estimated boundary using gradients, which makes exploring the change area faster and costs less computing time. Their approach works better with fewer samples compared to older methods, helping model complex systems more effectively.
What this means in practice
- •For simulation developers: Identify exact parameter regions where simulation outputs change abruptly to improve localized modeling and reduce training data needs.
- •For engineering teams: Detect operational regime shifts in numerical simulations to optimize design processes relying on complex models with sudden behavior changes.
Authors
Federico Betti, Eleni Chatzi, Paolo Tiso
Abstract
Many computational models exhibit multiple operating regimes separated by abrupt or non-smooth transitions as the input parameters vary. Such regime transitions complicate the construction of global parametric reduced-order models, often necessitating large training datasets to capture the distinct system behaviors. Accurate regime boundary detection instead enables localized modeling strategies by partitioning the parameter space into regions of qualitatively distinct behavior. In many cases, regime switching corresponds to the excursion set of a simulator output. When such a quantity is discontinuous or non-smooth, regression-based algorithms for level set estimation can struggle to correctly approximate the boundary. In this work, we show that solving a resulting binary classification problem can improve sample efficiency and we further propose a novel adaptive sampling strategy for support vector machines. Unlike uncertainty-based sampling schemes, the proposed approach uses a continuous search space and efficiently explores the currently estimated boundary in its tangential directions by gradient-based dynamic sampling. Because the decision function admits analytical gradients, this exploration incurs negligible computational overhead relative to the underlying model evaluations. Then, a space-filling subset of the generated samples is selected as the next batch to be queried. We demonstrate improved sample efficiency compared to existing active learning algorithms on analytical and numerical examples. The proposed method provides a sample-efficient framework for regime boundary detection in a broad class of discontinuous simulation-driven problems.