analiza aiqbench
This paper reframes diffusion-based generative modeling through iterative Gaussian posterior inference, introducing Bayesian Sample Inference (BSI) as a unified framework encompassing existing methods like BFNs and Variational Diffusion Models. Experimental results on ImageNet demonstrate that the proposed approach yields superior sample quality while maintaining competitive log-likelihood scores compared to current state-of-the-art baselines.
Generative Modeling with Bayesian Sample Inference
arXiv:2502.07580v4 Announce Type: replace Abstract: We present a novel view of diffusion-like generative modeling from the perspective of iterative Gaussian posterior inference. By treating the generated sample as an unknown variable, we formulate the sampling process in the language of Bayesian probability: at each step, a model predicts the unknown sample from our current belief state and we compute a posterior belief from that prediction. Based on this formulation, we propose the generative model Bayesian Sample Inference (BSI). In addition to a rigorous theoretical analysis, we show that our perspective includes Bayesian Flow Networks (BFNs) and that BSI and BFNs represent the two opposite ends of a range of isotropic hyper-priors over the initial belief state. In our experiments, we demonstrate that BSI improves sample quality over both BFNs and the closely related Variational Diffusion Models on ImageNet32 and ImageNet64, while achieving equivalent log-likelihoods on both datasets.
Abstract:We present a novel view of diffusion-like generative modeling from the perspective of iterative Gaussian posterior inference. By treating the generated sample as an unknown variable, we formulate the sampling process in the language of Bayesian probability: at each step, a model predicts the unknown sample from our current belief state and we compute a posterior belief from that prediction. Based on this formulation, we propose the generative model Bayesian Sample Inference (BSI). In addition to a rigorous theoretical analysis, we show that our perspective includes Bayesian Flow Networks (BFNs) and that BSI and BFNs represent the two opposite ends of a range of isotropic hyper-priors over the initial belief state. In our experiments, we demonstrate that BSI improves sample quality over both BFNs and the closely related Variational Diffusion Models on ImageNet32 and ImageNet64, while achieving equivalent log-likelihoods on both datasets.
Machine Learning (cs.LG); Machine Learning (stat.ML)
arXiv:2502.07580 [cs.LG]
(or arXiv:2502.07580v4 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2502.07580 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 237 cuvinte
Subiecte conexe
Foundation Models for Oversight
LessWrong · 28 iul. 2026
BayesNDE: Bayesian Generative Modeling for Neural Density Estimation
arXiv · 1 oct. 2026
Generative Bayesian Filtering for State Estimation
arXiv · 24 iul. 2026
Joint Bayesian Inference of Graphical Structure and Parameters with a Single Generative Flow Network
arXiv · 31 aug. 2026