Mean Field Variational

Mean-field variational Bayes (MFVB) is a computationally efficient approximation technique for Bayesian inference, aiming to simplify complex posterior distributions by assuming independence between variables. Current research focuses on improving the accuracy and scalability of MFVB, particularly within Bayesian neural networks (BNNs), exploring variations like particle-based MFVB and addressing challenges such as calibration of uncertainty estimates and the behavior of MFVB in overparameterized models. These advancements are significant for enabling Bayesian methods in high-dimensional problems, leading to more robust uncertainty quantification and improved reliability in applications ranging from particle physics to statistical regression.

Papers

August 1, 2024

Calibrating Bayesian Generative Machine Learning for Bayesiamplification
Sebastian Bieringer, Sascha Diefenbacher, Gregor Kasieczka, Mathias Trabs
Generative Modeling Bayesian Perspective Bayes Rule Bayesian Uncertainty Distributional Uncertainty Mean Field Variational

June 18, 2024

A variational Bayes approach to debiased inference for low-dimensional parameters in high-dimensional linear regression
Ismaël Castillo, Alice L'Huillier, Kolyan Ray, Luke Travis
High Dimensional Scientific Inference Low Rank Mean Field Sparse Linear Regression Parameter Space Variational Bayes Scalable Variational Mean Field Variational

April 11, 2023

Black Box Variational Inference with a Deterministic Objective: Faster, More Accurate, and Even More Black Box
Ryan Giordano, Martin Ingram, Tamara Broderick
Variational Inference Black Box Deterministic Algorithm Posterior Approximation Variational Approximation Black Box Variational Inference Posterior Covariance Mean Field Variational

March 24, 2023

Particle Mean Field Variational Bayes
Minh-Ngoc Tran, Paco Tseng, Robert Kohn
Bayesian Inference Langevin Dynamic Wasserstein Gradient Mean Field Variational

February 25, 2023

Bayesian Neural Networks Avoid Encoding Complex and Perturbation-Sensitive Concepts
Qihan Ren, Huiqi Deng, Yunuo Chen, Siyu Lou, Quanshi Zhang
Bayesian Neural Network Standard Deep Encoded Concept Mean Field Variational

July 8, 2022

Variational Inference of overparameterized Bayesian Neural Networks: a theoretical and empirical study
Tom Huix, Szymon Majewski, Alain Durmus, Eric Moulines, Anna Korba
Empirical Study Variational Inference Bayesian Neural Network Variational Posterior Mean Field Variational

February 23, 2022

Wide Mean-Field Bayesian Neural Networks Ignore the Data
Beau Coker, Wessel P. Bruinsma, David R. Burt, Weiwei Pan, Finale Doshi-Velez
Raw Data Bayesian Neural Network Mean Field Posterior Approximation Variational Density Mean Field Variational

Mean Field Variational

Papers

Calibrating Bayesian Generative Machine Learning for Bayesiamplification

A variational Bayes approach to debiased inference for low-dimensional parameters in high-dimensional linear regression

Black Box Variational Inference with a Deterministic Objective: Faster, More Accurate, and Even More Black Box

Particle Mean Field Variational Bayes

Bayesian Neural Networks Avoid Encoding Complex and Perturbation-Sensitive Concepts

Variational Inference of overparameterized Bayesian Neural Networks: a theoretical and empirical study

Wide Mean-Field Bayesian Neural Networks Ignore the Data