Search papers, labs, and topics across Lattice.
This paper formalizes the estimation of valid transport maps within a minimax framework, revealing that the sample complexity of learning any transport map is fundamentally linked to that of optimal transport maps under certain stability assumptions. The authors demonstrate that while estimating optimal transport maps is statistically challenging, alternative transport maps can be learned with greater accuracy when these assumptions do not hold. This work provides crucial insights into the statistical limits of generative modeling methods that rely on transport maps, such as diffusion models and flow matching.
Estimating valid transport maps can be as hard as optimal transport, but under certain conditions, alternative maps can be learned with significantly higher accuracy.
Many modern generative modeling methods, including diffusion models, normalizing flows, and flow matching, estimate transport maps or plans between distributions without explicitly targeting an optimal transport (OT) map. In applications like generative modeling, the transport cost itself is irrelevant, and this makes it natural to target maps which are more tractable from either a statistical or computational standpoint. In this short note, we formalize the task of estimating any valid transport map in a rigorous minimax framework. One consequence of this framing is that it yields sample complexity lower bounds for any method whose learned object is evaluated as a transport map or plan, including flow matching and diffusion-based generative models, in settings where direct analysis would be challenging due to the analytic complexity of the methods and their target maps. We observe that, under standard, though strong, stability assumptions from the OT literature, estimating any valid transport map is statistically as hard as estimating the OT map. We complement these results with some examples showing that when these stability assumptions fail, alternative transport maps can be learned substantially more accurately than the OT map. Our minimax framing provides a rigorous foundation for understanding the statistical limits of modern transport-based generative methods and clarifies when targeting sub-optimal maps can provide real statistical advantages.
CMU ML