Abstract:A class of generative models that unifies flow-based and diffusion-based methods is introduced. These models extend the framework proposed in Albergo & Vanden-Eijnden (2023), enabling the use of a broad class of continuous-time stochastic processes called `stochastic interpolants' to bridge any two arbitrary probability density functions exactly in finite time. These interpolants are built by combining data from the two prescribed densities with an additional latent variable that shapes the bridge in a flexible way. The time-dependent probability density function of the stochastic interpolant is shown to satisfy a first-order transport equation as well as a family of forward and backward Fokker-Planck equations with tunable diffusion coefficient. Upon consideration of the time evolution of an individual sample, this viewpoint immediately leads to both deterministic and stochastic generative models based on probability flow equations or stochastic differential equations with an adjustable level of noise. The drift coefficients entering these models are time-dependent velocity fields characterized as the unique minimizers of simple quadratic objective functions, one of which is a new objective for the score of the interpolant density. We show that minimization of these quadratic objectives leads to control of the likelihood for generative models built upon stochastic dynamics, while likelihood control for deterministic dynamics is more stringent. We also discuss connections with other methods such as score-based diffusion models, stochastic localization processes, probabilistic denoising techniques, and rectifying flows. In addition, we demonstrate that stochastic interpolants recover the Schrödinger bridge between the two target densities when explicitly optimizing over the interpolant. Finally, algorithmic aspects are discussed and the approach is illustrated on numerical examples.

RAP-modulated Fluid Processes: First Passages and the Stationary Distribution

First Exit Time of Markov Processes for Particle Transport in Fluidized Bed Reactors

A Stochastic Fluid Model Approach to the Stationary Distribution of the Maximum Priority Process

On the Existence and Uniqueness of Stationary Distributions for Some Piecewise Deterministic Markov Processes with State-Dependent Jump Intensity

Analytic and Monte Carlo Approximations to the Distribution of the First Passage Time of the Drifted Diffusion with Stochastic Resetting and Mixed Boundary Conditions

First Passage Times for Markov Renewal Processes and Applications

Regular stochastic flow and Dynamic Programming Principle for jump diffusions

A First Look at First-Passage Processes

A Computational Approach to First Passage Problems of Reflected Hyper-Exponential Jump Diffusion Processes

Duration-dependent stochastic fluid processes and solar energy revenue modeling

Mean Residence Time of Markov Processes for Particle Transport in Fludized Bed Reactors

Markov modulated fluid network process: Tail asymptotics of the stationary distribution

On Markov-dependent reflected autoregressive processes

Block-Structured Fluid Queues Driven by Qbd Processes

Computing first passage times for Markov-modulated fluid models using numerical PDE problem solvers

Bridging the Gap: A Generalized Stochastic Process for Count Data

First passage time statistics of Brownian motion with purely time dependent drift and diffusion

Diffusion Approximations for Self-excited Systems with Applications to General Branching Processes

Stochastic Interpolants: A Unifying Framework for Flows and Diffusions

Stochastic Suspended Sediment Dynamics in Semi-Bounded Open Channel Flows: A Reflected SDE Approach