Learning Robust Classifiers with Self-Guided Spurious Correlation Mitigation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zheng, Guangtao, Ye, Wenqian, Zhang, Aidong
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929336529977344
author Zheng, Guangtao
Ye, Wenqian
Zhang, Aidong
author_facet Zheng, Guangtao
Ye, Wenqian
Zhang, Aidong
contents Deep neural classifiers tend to rely on spurious correlations between spurious attributes of inputs and targets to make predictions, which could jeopardize their generalization capability. Training classifiers robust to spurious correlations typically relies on annotations of spurious correlations in data, which are often expensive to get. In this paper, we tackle an annotation-free setting and propose a self-guided spurious correlation mitigation framework. Our framework automatically constructs fine-grained training labels tailored for a classifier obtained with empirical risk minimization to improve its robustness against spurious correlations. The fine-grained training labels are formulated with different prediction behaviors of the classifier identified in a novel spuriousness embedding space. We construct the space with automatically detected conceptual attributes and a novel spuriousness metric which measures how likely a class-attribute correlation is exploited for predictions. We demonstrate that training the classifier to distinguish different prediction behaviors reduces its reliance on spurious correlations without knowing them a priori and outperforms prior methods on five real-world datasets.
format Preprint
id arxiv_https___arxiv_org_abs_2405_03649
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Learning Robust Classifiers with Self-Guided Spurious Correlation Mitigation
Zheng, Guangtao
Ye, Wenqian
Zhang, Aidong
Machine Learning
Computer Vision and Pattern Recognition
Deep neural classifiers tend to rely on spurious correlations between spurious attributes of inputs and targets to make predictions, which could jeopardize their generalization capability. Training classifiers robust to spurious correlations typically relies on annotations of spurious correlations in data, which are often expensive to get. In this paper, we tackle an annotation-free setting and propose a self-guided spurious correlation mitigation framework. Our framework automatically constructs fine-grained training labels tailored for a classifier obtained with empirical risk minimization to improve its robustness against spurious correlations. The fine-grained training labels are formulated with different prediction behaviors of the classifier identified in a novel spuriousness embedding space. We construct the space with automatically detected conceptual attributes and a novel spuriousness metric which measures how likely a class-attribute correlation is exploited for predictions. We demonstrate that training the classifier to distinguish different prediction behaviors reduces its reliance on spurious correlations without knowing them a priori and outperforms prior methods on five real-world datasets.
title Learning Robust Classifiers with Self-Guided Spurious Correlation Mitigation
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2405.03649