Paper Overview
Field: Machine Learning Author: Jiaming Li Published: 2026-08-17 arXiv: 2608.16873
Summary
High-fidelity finite-element simulations can provide accurate numerical predictions for side-branch resonators, but large simulation datasets are expensive to generate and purely data-driven surrogates may become unreliable when simulation-labelled data are scarce. This study develops an analytical-prior learning framework that reuses a low-cost analytical model to improve data efficiency under limited high-fidelity simulation budgets.
Key Points
- Problem: Simulation datasets for side-branch resonators are costly to generate; data-driven surrogates degrade when labelled data are scarce.
- Approach: An analytical-prior learning framework that leverages a low-cost analytical model, via two complementary routes:
- Explicit baseline: when the analytical model remains available at inference, retain it and learn only the analytical-to-simulation discrepancy from simulation data.
- Prior distillation: when a self-contained predictor is needed, first distill the analytical mapping from abundant low-cost evaluations into a learned prior, then calibrate it with limited simulation data.
- Evaluation: rectangular side-branch Helmholtz resonators, using 86 simulation-labelled geometries and 8,998 non-overlapping analytical-only geometries.
- Findings: analytical prior information significantly improves high-fidelity prediction when simulation data are scarce; explicit correction and prior distillation serve complementary deployment requirements.
Original Abstract
High-fidelity finite-element simulations can provide accurate numerical predictions for side-branch resonators, but large simulation datasets are expensive to generate and purely data-driven surrogates may become unreliable when simulation-labelled data are scarce. This study develops an analytical-prior learning framework that reuses a low-cost analytical model to improve data efficiency under limited high-fidelity simulation budgets. Two complementary routes are considered. When the analytical model remains available at inference, it is retained as an explicit baseline and the simulation data are used to learn only the analytical-to-simulation discrepancy. When a self-contained predictor is required, the analytical mapping is first distilled from abundant low-cost evaluations into a learned prior.
--- *Auto-collected on 2026-08-19*