Abstract:Flow matching (FM) is a family of training algorithms for fitting continuous normalizing flows (CNFs). A standard approach to FM, called conditional flow matching (CFM), exploits the fact that the marginal vector field of a CNF can be learned by fitting least-square regression to the so-called conditional vector field specified given one or both ends of the flow path. We show that viewing CFM training from a Bayesian decision theoretic perspective on parameter estimation opens the door to generalizations of CFM algorithms. We propose one such extension by introducing a CFM algorithm based on defining conditional probability paths given what we refer to as ``streams'', instances of latent stochastic paths that connect pairs of noise and observed data. Further, we advocate the modeling of these latent streams using Gaussian processes (GPs). The unique distributional properties of GPs, and in particular the fact that the velocity of a GP is still a GP, allows drawing samples from the resulting stream-augmented conditional probability path without simulating the actual streams, and hence the ``simulation-free" nature of CFM training is preserved. We show that this generalization of the CFM can substantially reduce the variance in the estimated marginal vector field at a moderate computational cost, thereby improving the quality of the generated samples under common metrics. Additionally, we show that adopting the GP on the streams allows for flexibly linking multiple related training data points (e.g., time series) and incorporating additional prior information. We empirically validate our claim through both simulations and applications to two hand-written image datasets.

Variational Flow Models: Flowing in Your Style

Mixed Variational Flows for Discrete Variables

VFlow: More Expressive Generative Flows with Variational Data Augmentation

Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models

Latent Variable Modelling with Hyperbolic Normalizing Flows

Variational Flow Matching for Graph Generation

β-Variational autoencoders and transformers for reduced-order modelling of fluid flows

MetFlow: A New Efficient Method for Bridging the Gap between Markov Chain Monte Carlo and Variational Inference

Conditional Variable Flow Matching: Transforming Conditional Densities with Amortized Conditional Optimal Transport

Improved variational inference with inverse autoregressive flow

Stochastic Sampling from Deterministic Flow Models

On Variational Methods For Fluid Flow Estimation

Automatic variational inference with cascading flows

Stream-level flow matching from a Bayesian decision theoretic perspective

Sinusoidal Flow: A Fast Invertible Autoregressive Flow

Normalizing Flow with Variational Latent Representation

$β$-Variational autoencoders and transformers for reduced-order modelling of fluid flows

An Adversarial Non-Volume Preserving Flow Model with Boltzmann Priors

Monge-Amp\`ere Flow for Generative Modeling

AmbientFlow: Invertible generative models from incomplete, noisy measurements