A Survey on Uncertainty Estimation in Deep Learning Classification Systems from a Bayesian Perspective JOSÉ MENA, Eurecat, Centre Tecnològic de Catalunya and Departament de Matemàtiques i Informàtica, Universitat de Barcelona, Spain ORIOL PUJOL, Departament de Matemàtiques i Informàtica, Universitat de Barcelona, Spain JORDI VITRIÀ, Departament de Matemàtiques i Informàtica, Universitat de Barcelona, Spain Decision-making based on machine learning systems, especially when this decision-making can afect human lives, is a subject of maximum interest in the Machine Learning community. It is, therefore, necessary to equip these systems with a means of estimating uncertainty in the predictions they emit in order to help practitioners make more informed decisions. In the present work, we introduce the topic of uncertainty estimation, and we analyze the peculiarities of such estimation when applied to classifcation systems. We analyze diferent methods that have been designed to provide classifcation systems based on deep learning with mechanisms for measuring the uncertainty of their predictions. We will take a look at how this uncertainty can be modeled and measured using diferent approaches, as well as practical considerations of diferent applications of uncertainty. Moreover, we review some of the properties that should be borne in mind when developing such metrics. All in all, the present survey aims to provide a pragmatic overview of the estimation of uncertainty in classifcation systems that can be very useful for both academic research and deep learning practitioners. CCS Concepts: · Computing methodologies → Neural networks; Supervised learning by classifca- tion;· Mathematics of computing → Bayesian networks. Additional Key Words and Phrases: Classifcation, Machine Learning, Deep Learning, Uncertainty ACM Reference Format: José Mena, Oriol Pujol, and Jordi Vitrià. 2022. A Survey on Uncertainty Estimation in Deep Learning Classif- cation Systems from a Bayesian Perspective. 1, 1 (February 2022), 36 pages. https://doi.org/10.1145/nnnnnnn. nnnnnnn 1 INTRODUCTION Machine learning (ML) is currently present in all kinds of applications and areas. Object recognition, automatic captioning, and machine translation represent just some of the multiple felds in which machine learning, and especially Deep Learning (DL), is imposing itself in the service of competitive business. In some areas of application, such as autonomous driving or automated patient diagnosis support systems, the level of performance required is very high. Failures in prediction can result in severe economic losses, or even the loss of human life. Hence the need for ways of managing the risks that automatic decision-making entails, and for these types of applications in particular. Authors’ addresses: José Mena, Eurecat, Centre Tecnològic de Catalunya, 72, Bilbao Street, Barcelona, Departament de Matemàtiques i Informàtica, Universitat de Barcelona, 585, Gran Via de les Corts Catalanes, Barcelona, Spain, jose.mena@ eurecat.org; Oriol Pujol, Departament de Matemàtiques i Informàtica, Universitat de Barcelona, 585, Gran Via de les Corts Catalanes, Barcelona, Spain, oriol_pujol@ub.edu; Jordi Vitrià, Departament de Matemàtiques i Informàtica, Universitat de Barcelona, 585, Gran Via de les Corts Catalanes, Barcelona, Spain, jordi.vitria@ub.edu. 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 proft or commercial advantage and that copies bear this notice and the full citation on the frst 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 specifc permission and/or a fee. Request permissions from permissions@acm.org. © 2022 Association for Computing Machinery. XXXX-XXXX/2022/2-ART $15.00 https://doi.org/10.1145/nnnnnnn.nnnnnnn , Vol. 1, No. 1, Article . Publication date: February 2022.