Data-Efficient Electromagnetic Surrogate Solver Through Dissipative Relaxation Transfer Learning

TL;DR

This paper introduces DIRTL, a transfer learning framework that enhances neural network surrogate models for electromagnetic simulations by stabilizing training around resonant phenomena, leading to improved accuracy and robustness.

Contribution

The paper presents a novel dissipative relaxation transfer learning method that pretrains models with a loss-induced smoothing technique, significantly improving electromagnetic surrogate solver performance.

Findings

Up to two-fold reduction in prediction error for FNO models.

Enhanced robustness and multi-tasking capabilities across training conditions.

Effective stabilization of high-resonance predictions through the DIRTL framework.

Abstract

In neural network surrogate solvers for electromagnetic simulations, accurately modeling resonant phenomena remains a central challenge. High-amplitude resonances generate strongly localized field patterns that deviate significantly from the general distribution of non-resonant cases, leading to instability and degraded predictive performance. To address this, we introduce dissipative relaxation transfer learning (DIRTL), a data-efficient training framework that integrates transfer learning with loss-regularized optimization principles from high-Q photonics. DIRTL first pretrains the model on data generated with a small fictitious material loss, which broadens sharp resonant modes and suppresses extreme field amplitudes. This smoothing of the response landscape enables the model to learn global modal features more effectively. The pretrained model is subsequently fine-tuned on the target lossless dataset containing true high-amplitude resonances, allowing stable adaptation based on the pretrained representation. Applied to both the Fourier Neural Operator (FNO) and UNet architectures, DIRTL yields substantial improvements in prediction accuracy, including up to a two-fold error reduction for the FNO variant. Furthermore, DIRTL demonstrates robustness across diverse training conditions and supports multi-tasking performance, suggesting the generalizability and flexibility of the pretrained core. Altogether, these results position DIRTL as a physically grounded curriculum for improving the reliability of neural network surrogate solvers.