Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper addresses the limitations of conditional diffusion models in emulating complex spatiotemporal dynamics by optimizing the noise schedule and training procedure. They characterize the relationship between noise schedule, reconstruction error, and exposure bias, revealing that standard schedules are suboptimal. Based on this, they introduce an Adaptive Noise Schedule framework to minimize reconstruction error and a Proxy Unrolled Training method to stabilize long-term rollouts, achieving significant improvements in accuracy and stability across various PDE benchmarks.
Diffusion models can achieve significantly higher accuracy and stability in PDE emulation by optimizing noise schedules to minimize reconstruction error and using a proxy unrolled training method, surpassing deterministic emulators.
Conditional Diffusion Models are powerful surrogates for emulating complex spatiotemporal dynamics, yet they often fail to match the accuracy of deterministic neural emulators for high-precision tasks. In this work, we address two critical limitations of autoregressive PDE diffusion models: their sub-optimal single-step accuracy and the prohibitive computational cost of unrolled training. First, we characterize the relationship between the noise schedule, the reconstruction error reduction rate and the diffusion exposure bias, demonstrating that standard schedules lead to suboptimal reconstruction error. Leveraging this insight, we propose an \textit{Adaptive Noise Schedule} framework that minimizes inference reconstruction error by dynamically constraining the model's exposure bias. We further show that this optimized schedule enables a fast \textit{Proxy Unrolled Training} method to stabilize long-term rollouts without the cost of full Markov Chain sampling. Both proposed methods enable significant improvements in short-term accuracy and long-term stability over diffusion and deterministic baselines on diverse benchmarks, including forced Navier-Stokes, Kuramoto-Sivashinsky and Transonic Flow.