Wasserstein gradient flows and forward diffusion struggle with multimodal sampling
Wasserstein Gradient Flows and Forward-Only Diffusion Are Not Enough for Multimodal Sampling
Machine Learning
Summary
Sampling complex distributions with multiple separate peaks is important but challenging. The authors show that popular methods based on Wasserstein gradient flows and forward-only diffusion face fundamental limits, often taking exponentially long times to explore all peaks. This is because these methods rely on local, gradient-based moves that struggle to jump between well-separated modes. Their work highlights that new, nonlocal methods are needed for efficiently sampling such complex distributions.
What this means in practice
- •For machine learning engineers: Avoid relying solely on Wasserstein gradient flows or forward diffusion when sampling from multimodal distributions, due to fundamental slow mixing limitations.
- •For computational chemists: Recognize the inherent challenges in applying local gradient-based sampling methods to explore well-separated molecular states efficiently.
A theory result. No direct application yet.
Authors
Daniel McBride, Pratik Khandagale, Cristina Garcia-Cardona, Yen Ting Lin
Abstract
There has been a proliferation of sampling algorithms based on Wasserstein gradient flows (WGF) and forward-only diffusion processes (FODP), often accompanied by theoretical guarantees of exponentially fast convergence to the target distribution. These guarantees are frequently interpreted as evidence that such methods can efficiently sample complex multimodal distributions, often supported by empirical results. In this work, we argue that this interpretation is fundamentally misleading. By invoking the Jordan-Kinderlehrer-Otto (JKO) scheme and Otto calculus, we establish that the canonical WGF sampling dynamics and overdamped forward diffusion share the same density evolution and therefore inherit the same metastability and slow-mixing phenomena long understood in nonequilibrium statistical physics. We analyze this family of samplers using two complementary tools -- spectral analysis and mean first-passage time (MFPT) analysis -- and show that well-separated multimodality can induce exponentially long mixing times associated with small spectral gaps and rare inter-mode transitions. For the commonly adopted log-linear annealing schedule studied here, we find that introducing intermediate distributions does not remove the exponential scaling of the total transport time. The limitation is structural rather than implementation-specific: purely local, gradient-driven transport mechanisms can require exponentially long times to transport probability mass across well-separated modes. We argue that this represents a fundamental limitation of WGF- and FODP-based sampling in their standard forms, and motivates future development of fundamentally nonlocal mechanisms for efficient multimodal sampling.