Vincent Fortuin
Fortuin coauthored Meta-Learning Mean Functions for Gaussian Processes. The study examines learning a prior mean from related tasks and discusses analytical evidence, practical benefits and the risk of overfitting. Bayesian Neural Network Priors Revisited was submitted in 2021 and accepted at ICLR 2022. Fortuin and collaborators studied weight correlations and heavy-tailed distributions, then tested priors based on those observations. Fortuin coauthored BNNpriors, a library supporting Markov Chain Monte Carlo inference with several prior families. The work provides modular tools to study uncertainty and the cold-posterior effect in Bayesian neural networks.
- Works
- 89
- Citations
- 698
- h-index
- 15
Publication numbers from OpenAlex, as of 14 October 2026.
Timeline
-
2019
Learning prior knowledge across tasks
Fortuin coauthored Meta-Learning Mean Functions for Gaussian Processes. The study examines learning a prior mean from related tasks and discusses analytical evidence, practical benefits and the risk of overfitting.
-
2021–2022
Examining neural-network priors
Bayesian Neural Network Priors Revisited was submitted in 2021 and accepted at ICLR 2022. Fortuin and collaborators studied weight correlations and heavy-tailed distributions, then tested priors based on those observations.
-
2021
Software for Bayesian inference
Fortuin coauthored BNNpriors, a library supporting Markov Chain Monte Carlo inference with several prior families. The work provides modular tools to study uncertainty and the cold-posterior effect in Bayesian neural networks.
Sources and credits
Photo: fortuin.github.io; Vincent Fortuin; creator not stated. License: Not stated by source. Source: fortuin.github.io. Cropped to a square and resized. Ask us to remove or replace this photo.