Neural operator benchmarks on 'The Well' dataset questioned by new study

A new arXiv paper challenges the performance benchmarks of neural operator models on 'The Well' dataset. The authors found that the original baselines, which were presented as illustrative and trained with minimal tuning, have been treated as definitive. By conducting a limited hyperparameter search for the Tucker-factorized Fourier Neural Operator (TFNO) under the same constraints, the researchers achieved significantly better results on most tasks, sometimes by orders of magnitude. This suggests that the initial neural operator baselines may not accurately reflect architectural performance. AI

IMPACT Highlights the need for rigorous evaluation and tuning in ML surrogate models for physical systems.

RANK_REASON The cluster contains an academic paper presenting new findings on model performance benchmarks. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• Fourier Neural Operator
• TFNO
• The WELL
• Tucker-factorized Fourier Neural Operator

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →