Organizations: Department of Applied Mathematics, The Hong Kong Polytechnic University, Hong Kong SAR, China · School of Mathematics and Statistics and Hubei Key Laboratory of Computational Science, Wuhan University, Wuhan, China · School of Mathematics and Statistics, Northeast Normal University, Changchun, China
We study distribution learning by flow matching with linear interpolation from a Gaussian source under Gaussian-channel regularity, a condition on the target distribution. For an integer r≥2, we establish spatial regularity of order r for the population velocity, together with temporal Lipschitz continuity on the full interval and local time analyticity before the terminal time. We use these properties to construct a finite multiscale ReLU sieve with a spatial Lipschitz bound independent of the truncation radius. For an exact minimizer of the empirical flow matching risk over this sieve, we obtain an upper bound on the expected Wasserstein-2 error of order n−r/(d+2r), up to a logarithmic factor, where n is the sample size and d is the dimension. The bound includes Euler discretization up to time one, without early stopping. The conditions hold at r=2 for targets with strongly convex C2 potentials and two-sided Hessian bounds. They also hold at every fixed order r≥2 for distributions obtained by convolving a law of bounded support with a nondegenerate isotropic Gaussian. For sufficiently large fixed constants defining the target class, we prove a minimax lower bound of order max{n−1/2,n−(r+1)/(d+2r)}.
Figures & tables
Condition
Consequence used in the proof
Real-tilt covariance bound, 0≤λ≤1
Centered posterior concentration, velocity derivatives on [0,1/2] , and target and response tails.
Linear growth of ∇ℓh
Linear velocity growth, the remainder in the posterior mean near t=1 , and the continuous endpoint value.
Bounds on hD2ℓh and hD3ℓh
Spatial Lipschitz and first-order time control, together with the covariance scale needed for complex continuation.
Bounds on hDmℓh , m≤r+1
Spatial Taylor accuracy of order r , hence the spatial approximation exponent.
Table 1: The roles of the conditions on the target distribution. All bounds are uniform over the specified target class.
Result
Polynomial rate
Guarantee
Kunkel (2026) , Theorem 4.3
n−2/(d+9+ϑ)
High-probability W1
Kunkel and Trabs (2026) , Remark 3.8(3)
n−2/(d+4)
Expected W1
Stéphanovitch et al. (2025) , Theorem I ( d≥3 )
n−3/(d+4)
Expected W1
Theorem 3.1 , r=2
n−2/(d+4)
Expected W2 and high-probability W2
Table 2: Guarantees on the common subclass ( H.1 ) for d≥2 . Logarithmic factors are suppressed and all class constants are fixed. Here ϑ>0 is fixed.
Flow-based generative modeling learns probability transports from simple noise to data for fast sampling, density estimation, and controllable generation.
Learning the underlying probability distribution from samples is a core problem in statistics and applied mathematics. Generative models based on diffusion and flow matching learn a velocity field whose flow transports a known distribution to the data distribution. They produce novel, realistic samples, although the empirical learning problem is ill posed near the data distribution. Its minimizer transports the samples onto the training data, and small perturbations of the data change the velocity field strongly near the end of the flow. The observed generalization is often attributed to implicit regularization by the network parameterization, early stopping of the training, and the time integration during sampling. Explicit regularization has advanced the theory and computation of many inverse problems and has recently been proposed to prevent memorization in generative models. With a regularizer on the magnitude of the velocity, however, the field approaches the memorizing velocity unless the regularization parameter grows to infinity fast enough, and it then stays bounded by a constant that is small only for large parameters. We show that penalizing the derivatives of order k of the velocity renders the D-dimensional empirical problem well posed uniformly in time if 2k>D and the training points are in general position. Our experiments in one and two dimensions discretize the velocity with splines and use direct solvers for the regularized least-squares problems to exclude optimization error. Refining the discretization in space and time changes the error of the Hessian-regularized flows only marginally, whereas without the regularizer the error tends to that of memorization. These flows generalize, with errors comparable to or lower than those of well-tuned kernel density estimates.
Lars Ruthotto, Eran Treister
Departments of Mathematics and Computer Science, Emory University, Atlanta, GA, USA · Faculty of Computer and Information Science, Ben-Gurion University of the Negev, Beer-Sheva, Israel
Despite the remarkable empirical success of flow-matching models, their statistical generalization guarantees remain underdeveloped. Existing analyses often impose restrictive assumptions on the estimated velocity field and yield convergence rates that fail to reflect the intrinsic low-dimensional structure common in real data, such as natural images and molecular geometries. In this work, we study the statistical generalization of flow-matching models for learning an unknown distribution Pdata from finitely many samples. We derive finite-sample error bounds on the learned generative distribution, measured in the Wasserstein-p distance, for all p≥1. Specifically, given n i.i.d. samples from Pdata, we show that, for every d>dp∗(Pdata) and appropriately chosen network architectures and hyperparameters, the learned distribution PFM satisfies Wp(PFM,Pdata)≲n−1/d+n−1/(2p)(log(1/ξ))1/(2p) with probability at least 1−ξ, where dp∗(Pdata) denotes the Wasserstein-p dimension of the target measure. Our results demonstrate that flow matching naturally adapts to the intrinsic geometry of data and mitigates the curse of dimensionality, as the convergence exponent depends on the intrinsic rather than ambient dimension. These guarantees remain meaningful in high-dimensional regimes and provide a theoretical explanation for the empirical success of flow matching on structured data distributions under substantially more relaxed assumptions than those in existing analyses.
Saptarshi Chakraborty, Quentin Berthet, Peter L. Bartlett
Department of Statistics, University of Michigan · Google DeepMind · Department of Statistics, University of California, Berkeley +1
Flow Matching (FM) learns a velocity field whose ODE transports a simple source distribution to a target law. Existing finite-sample theory largely treats ambient-space regularity or data supported exactly on low-dimensional sets. We study linear FM under a noisy latent-generator model, where a low-dimensional Hölder map is perturbed by nondegenerate ambient Gaussian noise, so the target law is full-dimensional despite its latent structure. We construct a spatially regular ReLU velocity class and establish non-asymptotic high-probability approximation and estimation bounds whose leading sample-size exponent is governed by the latent dimension rather than the ambient dimension, with ambient and noise dependence kept explicit. Fixed positive target noise keeps the interpolation nondegenerate over the full time interval. The same spatial regularity propagates the learned velocity error through the transport ODE, yielding a corresponding Wasserstein convergence guarantee. These results show that exact low-dimensional support is not necessary for Flow Matching to retain latent-dimensional statistical behavior.
Lifeng Hao, Shaolin Ji
Zhongtai Securities Institute for Financial Studies, Shandong University, Jinan 250100, Shandong, China.