Predictive methods improve flow matching under changing data distributions

Minimax-Optimal Predictive Flow Matching under Distribution Drift

Machine Learning

Summary

Flow matching trains models to generate data patterns by learning how these patterns move or flow. The challenge is that data can change over time, so the model must predict future patterns without seeing them yet. The authors propose a method called DyProd-FM that recombines past data in a clever way to adapt to these changing patterns. Their method achieves the best possible performance compared to other approaches, even when the underlying data or models are complex and non-linear.

What this means in practice

  • For machine learning engineers: Implement adaptive generative models that maintain accuracy when the data environment changes unpredictably over time.
  • For financial data analysts: Predict evolving market distributions more reliably by applying flow matching methods resilient to distribution shifts.

Authors

Aadithya Srikanth, Vaneet Aggarwal

Abstract

Flow matching learns a generative velocity field by regression, but under distribution drift a field must be deployed before the current data arrive. We study this predictive problem for a fixed, possibly nonconvex, neural class, with dynamic regret against the best represented field of each round. We propose DyProd-FM, which refits flow matching on dyadic historical prefixes and selects among the fits with a sleeping Prod master. Without knowing any drift budget, it attains regret $\widetilde{O}(\min\{N_ω+1,1+\sqrt{TΩ}\})$ in the relative velocity risk variation $Ω$ and $\widetilde{O}(\min\{N_μ+1,1+T^{1/3}V_μ^{2/3}\})$ in the target-law variation $V_μ=\sum_t W_2(μ_{t+1},μ_t)$, up to approximation and fitting costs. Matching lower bounds for fixed classes show that both rates are optimal in $T$ and in the budgets up to logarithmic factors. Prior optimal dynamic-regret guarantees rely on a convex decision set with strongly convex or exp-concave losses, whereas our class may be nonconvex and no curvature in its parameters is assumed. The target-law rate rests on a second-order link between the two variations, which the base noise of flow matching provides.