PDEfuncta: Spectrally-Aware Neural Representation for PDE Solution Modeling
Preprint
Abstract
Scientific machine learning often involves representing complex solution fields that exhibit high-frequency features such as sharp transitions, fine-scale oscillations, and localized structures. While implicit neural representations (INRs) have shown promise for continuous function modeling, capturing such high-frequency behavior remains a challenge-especially when modeling multiple solution fields with a shared network. Prior work addressing spectral bias in INRs has primarily focused on single-instance settings, limiting scalability and generalization. In this work, we propose Global Fourier Modulation (GFM), a novel modulation technique that injects high-frequency information at each layer of the INR through Fourier-based reparameterization. This enables compact and accurate representation of multiple solution fields using low-dimensional latent vectors. Building upon GFM, we introduce PDEfuncta, a meta-learning framework designed to learn multi-modal solution fields and support generalization to new tasks. Through empirical studies on diverse scientific problems, we demonstrate that our method not only improves representational quality but also shows potential for forward and inverse inference tasks without the need for retraining.Keywords
All Related Versions
This publication has 10 references indexed in Scilit:
- Neural operators for accelerating scientific simulations and designNature Reviews Physics, 2024
- Geometry-Informed Neural Operator for Large-Scale 3D PDEs2023
- Operator Learning with Neural Fields: Tackling PDEs on General Geometries2023
- Scientific Machine Learning Through Physics–Informed Neural Networks: Where we are and What’s NextJournal of Scientific Computing, 2022
- OpenFWI: Large-Scale Multi-Structural Benchmark Datasets for Full Waveform Inversion2022
- PDEBench: An Extensive Benchmark for Scientific Machine Learning2022
- Neural tangent kernel: convergence and generalization in neural networks (invited paper)2021
- Learning nonlinear operators via DeepONet based on the universal approximation theorem of operatorsNature Machine Intelligence, 2021
- Workshop Report on Basic Research Needs for Scientific Machine Learning: Core Technologies for Artificial Intelligence2019
- Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equationsJournal of Computational Physics, 2019