Abstract:Diffusion models are a new class of generative models that revolve around the estimation of the score function associated with a stochastic differential equation. Subsequent to its acquisition, the approximated score function is then harnessed to simulate the corresponding time-reversal process, ultimately enabling the generation of approximate data samples. Despite their evident practical significance these models carry, a notable challenge persists in the form of a lack of comprehensive quantitative results, especially in scenarios involving non-regular scores and estimators. In almost all reported bounds in Kullback Leibler (KL) divergence, it is assumed that either the score function or its approximation is Lipschitz uniformly in time. However, this condition is very restrictive in practice or appears to be difficult to establish. To circumvent this issue, previous works mainly focused on establishing convergence bounds in KL for an early stopped version of the diffusion model and a smoothed version of the data distribution, or assuming that the data distribution is supported on a compact manifold. These explorations have led to interesting bounds in either Wasserstein or Fortet-Mourier metrics. However, the question remains about the relevance of such early-stopping procedure or compactness conditions. In particular, if there exist a natural and mild condition ensuring explicit and sharp convergence bounds in KL. In this article, we tackle the aforementioned limitations by focusing on score diffusion models with fixed step size stemming from the Ornstein-Uhlenbeck semigroup and its kinetic counterpart. Our study provides a rigorous analysis, yielding simple, improved and sharp convergence bounds in KL applicable to any data distribution with finite Fisher information with respect to the standard Gaussian distribution.

Computing Nonequilibrium Responses with Score-shifted Stochastic Differential Equations

Stationary Response of Stochastically Excited Nonlinear Systems with Continuous-Time Markov Jump

Determining Response Probability Density of Nonlinear Stochastic Systems by Modifying Analytical Result of Degenerated Systems.

Stochastic Averaging of Strongly Nonlinear Oscillators Under Combined Harmonic and Wide-Band Noise Excitations

Response Theory via Generative Score Modeling

Linear response in large deviations theory: A method to compute non-equilibrium distributions

Closing the ODE-SDE gap in score-based diffusion models through the Fokker-Planck equation

Nonequilibrium Response for Markov Jump Processes: Exact Results and Tight Bounds

KL Convergence Guarantees for Score diffusion models under minimal data assumptions

Ergodic and foliated kernel-differentiation method for linear responses of random systems

Probabilistic Predictability of Stochastic Dynamical Systems

Score-based Diffusion Models via Stochastic Differential Equations -- a Technical Tutorial

A systematic path to non-Markovian dynamics II: Probabilistic response of nonlinear multidimensional systems to Gaussian colored noise excitation

Universal Non-equilibrium Response Theory Beyond Steady States

A statistical learning framework for mapping indirect measurements of ergodic systems to emergent properties

Fluctuation-response inequalities for kinetic and entropic perturbations

Large Deviation Minimisers for Stochastic Partial Differential Equations with Degenerate Noise

Score Dynamics: Scaling Molecular Dynamics with Picoseconds Time Steps via Conditional Diffusion Model

Fluctuations in Wasserstein dynamics on Graphs