Hands-on Bayesian Neural Networks - a Tutorial for Deep
Learning Users
LAURENT VALENTIN JOSPIN, University of Western Australia
WRAY BUNTINE, Monash University
FARID BOUSSAID, University of Western Australia
HAMID LAGA, Murdoch university
MOHAMMED BENNAMOUN, University of Western Australia
Modern deep learning methods have equipped researchers and engineers with incredibly powerful tools to
tackle problems that previously seemed impossible. However, since deep learning methods operate as black
boxes, the uncertainty associated with their predictions is often challenging to quantify. Bayesian statistics
oer a formalism to understand and quantify the uncertainty associated with deep neural networks predictions.
This paper provides a tutorial for researchers and scientists who are using machine learning, especially deep
learning, with an overview of the relevant literature and a complete toolset to design, implement, train, use
and evaluate Bayesian neural networks.
CCS Concepts:
• Mathematics of computing → Probability and statistics
;
• Computing methodologies
→ Neural networks; Bayesian network models; Ensemble methods; Regularization.
Additional Key Words and Phrases: Bayesian methods, Bayesian Deep Learning, Approximate Bayesian
methods
ACM Reference Format:
Laurent Valentin Jospin, Wray Buntine, Farid Boussaid, Hamid Laga, and Mohammed Bennamoun. 2020.
Hands-on Bayesian Neural Networks - a Tutorial for Deep Learning Users. ACM Comput. Surv. 1, 1 (July 2020),
35 pages. https://doi.org/10.1145/nnnnnnn.nnnnnnn
1 INTRODUCTION
Deep learning has led to a revolution in machine learning, providing solutions to tackle more and
more complex and challenging real-life problems. However, deep learning models are prone to
overtting, which adversely aects their generalization capabilities. Deep learning models also
tend to be overcondent about their predictions (when they do provide a condence interval). All
of this is problematic for applications such as self driving cars [
74
], medical diagnostics [
38
] or
trading and nance [
11
], where silent failure can lead to dramatic outcomes. Consequently, many
approaches have been proposed to mitigate this risk, especially via the use of stochastic neural
networks to estimate the uncertainty in the model prediction. The Bayesian paradigm provides a
Authors’ addresses: Laurent Valentin Jospin, laurent.jospin@research.uwa.edu.au, University of Western Australia, 35
Stirling Hwy, Crawley, Western Australia, 6009; Wray Buntine, wray.buntine@monash.edu, Monash University, Wellington
Rd, Monash, Victoria, 3800; Farid Boussaid, farid.boussaid@uwa.edu.au, University of Western Australia, 35 Stirling Hwy,
Crawley, Western Australia, 6009; Hamid Laga, h.laga@murdoch.edu.au, Murdoch university, 90 South St, Murdoch, Western
Australia, 6150; Mohammed Bennamoun, mohammed.bennamoun@uwa.edu.au, University of Western Australia, 35 Stirling
Hwy, Crawley, Western Australia, 6009.
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee
provided that copies are not made or distributed for prot or commercial advantage and that copies bear this notice and
the full citation on the rst page. Copyrights for components of this work owned by others than ACM must be honored.
Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires
prior specic permission and/or a fee. Request permissions from permissions@acm.org.
© 2020 Association for Computing Machinery.
0360-0300/2020/7-ART $15.00
https://doi.org/10.1145/nnnnnnn.nnnnnnn
ACM Comput. Surv., Vol. 1, No. 1, Article . Publication date: July 2020.
arXiv:2007.06823v1 [cs.LG] 14 Jul 2020