Papers/2608.24940
🧪 Test?View on arXiv

When Does Frequency Decomposition Benefit Physics-Informed Neural Networks? A Preliminary Ablation Study

Not provided in the content

PDEsneural networksfrequency decompositionspectral analysis
2608.24940
Builder Relevance
70%
3h ago

Abstract

This study explores the effectiveness of frequency decomposition in enhancing the performance of Physics-Informed Neural Networks (PINNs) on various partial differential equations (PDEs).

Reality Card

Core Claim

Frequency decomposition significantly improves the accuracy of PINNs on spectrally complex PDEs, reducing relative $L_2$ error by up to 59.2% on a multimodal wave problem.

Method / Result

Frequency decomposition helps most on spectrally complex benchmarks, achieving up to 59.2% reduction in $L_2$ error.

Limitations

All results are based on a single training seed across five benchmarks, raising concerns about reproducibility and generalizability.

Paper to code

Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.

No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.
← Back to all papers