Energy Based Prior

Energy-based priors (EBPs) are increasingly used in generative modeling to improve the expressiveness and informativeness of latent variable distributions, particularly in multimodal settings. Current research focuses on integrating EBMs with various generative models, such as diffusion models and hierarchical architectures, often employing Markov Chain Monte Carlo (MCMC) methods, including Langevin dynamics and diffusion-amortized MCMC, for efficient sampling. This approach addresses limitations of simpler priors like Gaussian distributions, leading to more coherent and accurate generation across different data modalities. The improved generative capabilities of EBPs have implications for various applications, including image generation, medical imaging reconstruction, and out-of-distribution detection.

Papers

September 30, 2024

Learning Multimodal Latent Generative Models with Energy-Based Prior
Shiyu Yuan, Jiali Cui, Hanao Li, Tian Han
Generative Model Multimodal Phenomenon Multimodal Generative Model Semantic Coherence Energy Based Prior

August 20, 2024

Learning Multimodal Latent Space with EBM Prior and MCMC Inference
Shiyu Yuan, Carlo Lipizzi, Tian Han
Generative Model Latent Space Multimodal Phenomenon Markov Chain Monte Carlo Multimodal Generation Joint Generative Energy Based Prior

May 22, 2024

Learning Latent Space Hierarchical EBM Diffusion Models
Jiali Cui, Tian Han
Latent Diffusion Prior Model Posterior Network Energy Based Prior

October 5, 2023

Learning Energy-Based Prior Model with Diffusion-Amortized MCMC
Peiyu Yu, Yaxuan Zhu, Sirui Xie, Xiaojian Ma, Ruiqi Gao, Song-Chun Zhu, Ying Nian Wu
LeArning Abstract Generative Modeling Markov Chain Monte Carlo Posterior Sampling Energy Based Prior Model Energy Based Prior

January 28, 2023

Master's Thesis: Out-of-distribution Detection with Energy-based Models
Sven Elflein
Deep Learning Adversarial Example Distribution Detection Deep Network Expert Knowledge Energy Based Model Electronic Thesis Training Data Distribution Energy Based Prior

April 19, 2022

An Energy-Based Prior for Generative Saliency
Jing Zhang, Jianwen Xie, Nick Barnes, Ping Li
Accurate Saliency Energy Based Prior

March 23, 2022

Computed Tomography Reconstruction using Generative Energy-Based Priors
Martin Zach, Erich Kobler, Thomas Pock
Generative Question Computed Tomography Reconstruction Reconstruction Algorithm Robust Reconstruction Variational Regularization Energy Based Prior