New latent PDE mapping technique enables geometry generalization in physics-informed learning

Researchers have developed a new physics-informed learning technique called latent PDE mapping, which allows machine learning models to generalize across different geometries using limited training data. This method maps geometry-specific PDE residuals and boundary conditions to a latent geometry, enabling the calculation of geometry-consistent shape gradients. The technique was demonstrated by solving the Aliev-Panfilov PDE in cardiac electrophysiology with both physics-informed neural networks and deep operator networks, achieving a significant reduction in error with minimal computational cost during inference. AI

IMPACT This technique could significantly improve the efficiency and generalizability of physics-informed machine learning models, particularly in fields with complex geometries and limited data.

RANK_REASON The cluster contains a research paper detailing a new method for physics-informed learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Aliev-Panfilov PDE
• arXiv
• cardiac electrophysiology
• cs.LG
• Ingvild Askim Adde
• Latent PDE mapping
• physics-informed deep operator networks
• physics-informed neural networks

• paper
• model release

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