Functional Prior

Functional priors are increasingly used in Bayesian machine learning to incorporate prior knowledge about the shape or properties of a function into model training, improving efficiency and accuracy. Current research focuses on integrating functional priors with various models, including Bayesian neural networks and Gaussian processes, often employing techniques like anchored ensembling, amortized active learning, and Laplace approximations for efficient inference. This approach enhances the performance of Bayesian optimization, active learning, and uncertainty quantification in diverse applications such as materials science, time series analysis, and personalized medicine, particularly where data is scarce or expensive to acquire. The resulting improved model accuracy and uncertainty estimates lead to more reliable predictions and better decision-making in these fields.

Papers

October 21, 2024

High-Fidelity Transfer of Functional Priors for Wide Bayesian Neural Networks by Learning Activations
Marcin Sendera, Amin Sorkhei, Tomasz Kuśmierczyk
Network Programming Bayesian Neural Network Wide Neural Network Functional Prior Complex Posterior

September 8, 2024

Empowering Bayesian Neural Networks with Functional Priors through Anchored Ensembling for Mechanics Surrogate Modeling Applications
Javad Ghorbanian, Nicholas Casaprima, Audrey Olivier
Bayesian Neural Network Model Ensembling Neural Network Weight Fidelity Model Surrogate Modelling Functional Prior

July 25, 2024

Amortized Active Learning for Nonparametric Functions
Cen-You Li, Marc Toussaint, Barbara Rakitsch, Christoph Zimmer
Active Learning Nonparametric Regression Functional Prior Active Causal Active SEt

July 18, 2024

FSP-Laplace: Function-Space Priors for the Laplace Approximation in Bayesian Deep Learning
Tristan Cinquin, Marvin Pförtner, Vincent Fortuin, Philipp Hennig, Robert Bamler
Inductive Bias Bayesian Deep Learning Laplace Approximation Gaussian Prior Functional Prior Laplace fPINNs

May 29, 2024

Efficient Black-box Adversarial Attacks via Bayesian Optimization Guided by a Function Prior
Shuyu Cheng, Yibo Miao, Yinpeng Dong, Xiao Yang, Xiao-Shan Gao, Jun Zhu
Adversarial Attack Adversarial Example Bayesian Optimization Black Box Model Black Box Adversarial Attack Functional Prior

November 8, 2023

Functional Bayesian Tucker Decomposition for Continuous-indexed Tensor Data
Shikai Fang, Xin Yu, Zheng Wang, Shibo Li, Mike Kirby, Shandian Zhe
Tensor Data Tucker Decomposition Tensor Model Functional Prior

August 28, 2023

BayOTIDE: Bayesian Online Multivariate Time series Imputation with functional decomposition
Shikai Fang, Qingsong Wen, Yingtao Luo, Shandian Zhe, Liang Sun
Gaussian Process Time Series Data Time Series Imputation Function Decomposition Functional Prior

July 7, 2022

Pre-training helps Bayesian optimization too
Zi Wang, George E. Dahl, Kevin Swersky, Chansoo Lee, Zelda Mariet, Zachary Nado, Justin Gilmer, Jasper Snoek, Zoubin Ghahramani
Bayesian Optimization Pre Training Global Optimization Functional Prior

June 8, 2022

Ensembles for Uncertainty Estimation: Benefits of Prior Functions and Bootstrapping
Vikranth Dwaracherla, Zheng Wen, Ian Osband, Xiuyuan Lu, Seyed Mohammad Asghari, Benjamin Van Roy
Uncertainty Estimation Ensemble Learning Diverse Ensemble Complementary Benefit Bootstrapping End to End Ensemble Framework Functional Prior

May 5, 2022

A Deep Bayesian Bandits Approach for Anticancer Therapy: Exploration via Functional Prior
Mingyu Lu, Yifang Chen, Su-In Lee
Environment Exploration Contextual Bandit Drug Discovery Drug Feature Drug Response Prediction Precision Oncology Neural Bandit Functional Prior

Functional Prior

Papers

High-Fidelity Transfer of Functional Priors for Wide Bayesian Neural Networks by Learning Activations

Empowering Bayesian Neural Networks with Functional Priors through Anchored Ensembling for Mechanics Surrogate Modeling Applications

Amortized Active Learning for Nonparametric Functions

FSP-Laplace: Function-Space Priors for the Laplace Approximation in Bayesian Deep Learning

Efficient Black-box Adversarial Attacks via Bayesian Optimization Guided by a Function Prior

Functional Bayesian Tucker Decomposition for Continuous-indexed Tensor Data

BayOTIDE: Bayesian Online Multivariate Time series Imputation with functional decomposition

Pre-training helps Bayesian optimization too

Ensembles for Uncertainty Estimation: Benefits of Prior Functions and Bootstrapping

A Deep Bayesian Bandits Approach for Anticancer Therapy: Exploration via Functional Prior