New FAST-DeepONet method enhances AI stability for complex PDE problems

Researchers have developed FAST-DeepONet, a novel approach to improve the statistical stability of Deep Operator Networks when dealing with high-dimensional inputs from partial differential equations (PDEs). This new method combines a fixed spectral path with a regularized projection of the orthogonal residual, effectively mitigating performance degradation even with a large number of correlated sensors and limited operator samples. FAST-DeepONet demonstrates significant error reduction and parameter efficiency across various applications, including Navier--Stokes flow, Darcy flow, and terminal wavefield prediction. AI

IMPACT Improves AI model performance and stability for complex scientific simulations.

RANK_REASON Academic paper detailing a new method for AI models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• DeepONet
• FAST-DeepONet
• Navier--Stokes flow

• paper
• model release

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