Beyond Loss Guidance: Using PDE Residuals as Spectral Attention in Diffusion Neural Operators
This addresses inefficiencies in PDE solving for scientific computing, offering a faster and more robust alternative to existing methods, though it is incremental as it builds on diffusion neural operators.
The paper tackled the problem of slow and unstable gradient-based optimization in diffusion-based PDE solvers by introducing PRISMA, a conditional diffusion neural operator that embeds PDE residuals directly into the architecture via spectral attention, achieving competitive accuracy with 15x to 250x faster inference and 10x to 100x fewer denoising steps across five benchmark PDEs.
Diffusion-based solvers for partial differential equations (PDEs) are often bottle-necked by slow gradient-based test-time optimization routines that use PDE residuals for loss guidance. They additionally suffer from optimization instabilities and are unable to dynamically adapt their inference scheme in the presence of noisy PDE residuals. To address these limitations, we introduce PRISMA (PDE Residual Informed Spectral Modulation with Attention), a conditional diffusion neural operator that embeds PDE residuals directly into the model's architecture via attention mechanisms in the spectral domain, enabling gradient-descent free inference. In contrast to previous methods that use PDE loss solely as external optimization targets, PRISMA integrates PDE residuals as integral architectural features, making it inherently fast, robust, accurate, and free from sensitive hyperparameter tuning. We show that PRISMA has competitive accuracy, at substantially lower inference costs, compared to previous methods across five benchmark PDEs, especially with noisy observations, while using 10x to 100x fewer denoising steps, leading to 15x to 250x faster inference.