Posterior Mode

Posterior mode estimation focuses on accurately characterizing the probability distribution of model parameters, particularly in complex models like Bayesian neural networks, aiming for robust and reliable predictions. Current research emphasizes improving posterior approximation methods, including exploring the impact of posterior flatness and multimodality, and developing efficient algorithms like variational inference, Hamiltonian Monte Carlo, and Gibbs sampling, often within the context of deep learning architectures. These advancements are crucial for enhancing the accuracy and reliability of predictions in various applications, from image analysis and astroparticle physics to natural language processing and reinforcement learning, by providing better uncertainty quantification and improved model generalization.

Papers

June 21, 2023

Beyond Deep Ensembles: A Large-Scale Evaluation of Bayesian Deep Learning under Distribution Shift
Florian Seligmann, Philipp Becker, Michael Volpp, Gerhard Neumann
Distribution Shift Deep Ensemble Large Pre Trained Model Bayesian Deep Learning Approximate Inference Posterior Mode Large Scale Evaluation

June 13, 2023

DORSal: Diffusion for Object-centric Representations of Scenes et al
Allan Jabri, Sjoerd van Steenkiste, Emiel Hoogeboom, Mehdi S. M. Sajjadi, Thomas Kipf
Diffusion Explainer Neural Rendering Object Centric Representation 3D Scene Representation Posterior Mode 3D Scene Generation Slot Representation Object Scene

June 5, 2023

Gibbs Sampling the Posterior of Neural Networks
Giovanni Piccioli, Emanuele Troiani, Lenka Zdeborová
Neural Network Langevin Dynamic Markov Chain Monte Carlo Hamiltonian Monte Carlo Posterior Mode Gibbs Algorithm Gibbs Sampler Reversible Markov Chain

April 27, 2023

Variational Bayes Made Easy
Mohammad Emtiyaz Khan
Approximate Inference Posterior Mode Variational Bayes Mathematical Derivation Multiple Distribution

March 14, 2023

Graph Neural Network Surrogates of Fair Graph Filtering
Emmanouil Krasanakis, Symeon Papadopoulos
Recommender System Universal Approximation Graph Filter Posterior Mode Graph Mining Task

December 30, 2022

Posterior sampling with CNN-based, Plug-and-Play regularization with applications to Post-Stack Seismic Inversion
Muhammad Izzatullah, Tariq Alkhalifah, Juan Romero, Miguel Corrales, Nick Luiken, Matteo Ravasi
Financial Application Variational Inference Regularization Model Ill Posed Inverse Problem Seismic Data Posterior Mode Bayesian Inversion Seismic Inversion

October 16, 2022

Posterior Regularized Bayesian Neural Network Incorporating Soft and Hard Knowledge Constraints
Jiayu Huang, Yutian Pang, Yongming Liu, Hao Yan
Uncertainty Quantification Bayesian Neural Network Posterior Mode Neural Network Parameter Bayesian Regularization

May 30, 2022

Posterior and Computational Uncertainty in Gaussian Processes
Jonathan Wenger, Geoff Pleiss, Marvin Pförtner, Philipp Hennig, John P. Cunningham
Gaussian Process Covariance Conditioning Posterior Mode Computational Uncertainty Posterior Covariance

May 23, 2022

Split personalities in Bayesian Neural Networks: the case for full marginalisation
David Yallup, Will Handley, Mike Hobson, Anthony Lasenby, Pablo Lemos
Bayesian Neural Network Case Relevance Posterior Mode Multiple Mind

May 19, 2022

Foundation Posteriors for Approximate Probabilistic Inference
Mike Wu, Noah Goodman
Zero Shot Posterior Inference Posterior Mode Variational Objective Probabilistic Program

December 7, 2021

On the Effectiveness of Mode Exploration in Bayesian Model Averaging for Neural Networks
John T. Holodnak, Allan B. Wollaber
Neural Network Variational Inference Deep Ensemble Posterior Mode Bayesian Model Averaging Mode Matching