Striking the right balance with uncertainty

Salman Khan, Munawar HAYAT, Syed Waqas Zamir, Jianbing Shen, Ling Shao

Research output: A Conference proceeding or a Chapter in BookConference contributionpeer-review

128 Citations (Scopus)

Abstract

Learning unbiased models on imbalanced datasets is a significant challenge. Rare classes tend to get a concentrated
representation in the classification space which hampers the generalization of learned boundaries to new test
examples. In this paper, we demonstrate that the Bayesian uncertainty estimates directly correlate with the rarity of
classes and the difficulty level of individual samples. Subsequently, we present a novel framework for uncertainty
based class imbalance learning that follows two key insights: First, classification boundaries should be extended
further away from a more uncertain (rare) class to avoid over-fitting and enhance its generalization. Second, each
sample should be modeled as a multi-variate Gaussian distribution with a mean vector and a covariance matrix defined
by the sample’s uncertainty. The learned boundaries should respect not only the individual samples but also their
distribution in the feature space. Our proposed approach efficiently utilizes sample and class uncertainty information
to learn robust features and more generalizable classifiers.
We systematically study the class imbalance problem and derive a novel loss formulation for max-margin learning
based on Bayesian uncertainty measure. The proposed method shows significant performance improvements on six
benchmark datasets for face verification, attribute prediction, digit/object classification and skin lesion detection.
Original languageEnglish
Title of host publicationProceedings - 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2019
PublisherIEEE
Pages103-112
Number of pages10
ISBN (Electronic)9781728132938
DOIs
Publication statusPublished - Jun 2019
Externally publishedYes
EventIEEE Conference on Computer Vision and Pattern Recognition - Long Beach, United States
Duration: 15 Jun 201920 Jun 2019
http://cvpr2019.thecvf.com/

Publication series

NameProceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
Volume2019-June
ISSN (Print)1063-6919

Conference

ConferenceIEEE Conference on Computer Vision and Pattern Recognition
Abbreviated titleCVPR 2019
Country/TerritoryUnited States
CityLong Beach
Period15/06/1920/06/19
Internet address

Fingerprint

Dive into the research topics of 'Striking the right balance with uncertainty'. Together they form a unique fingerprint.

Cite this