Debiased Learning via Composed Conceptual Sensitivity Regularization

Deep neural networks often rely on spurious features, which are attributes correlated with class labels but irrelevant to the actual task, leading to poor generalization when these features are absent. To train classifiers that are not biased towards spurious features, recent research has leveraged...

Full description

Saved in:
Bibliographic Details
Main Authors: Sunghwan Joo, Taesup Moon
Format: Article
Language:English
Published: IEEE 2024-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10713328/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846160755540361216
author Sunghwan Joo
Taesup Moon
author_facet Sunghwan Joo
Taesup Moon
author_sort Sunghwan Joo
collection DOAJ
description Deep neural networks often rely on spurious features, which are attributes correlated with class labels but irrelevant to the actual task, leading to poor generalization when these features are absent. To train classifiers that are not biased towards spurious features, recent research has leveraged explainable AI (XAI) techniques to identify and modify model behavior. Specifically, Concept Activation Vectors (CAVs), which indicate the direction toward specific concepts in the embedding space, were used to measure and regularize the conceptual sensitivity of the classifier, thereby reducing its reliance on spurious features. However, these approaches struggle with non-linear or high-dimensional spurious correlations due to the use of linear CAVs in previous works. In this paper, we propose Composite Conceptual Sensitivity Regularization (CCSR), a novel method designed to address these limitations. CCSR utilizes concept gradients to assign individualized CAVs for each sample, enabling the handling of non-linearly distributed spurious features in embedding space. Additionally, our method employs multiple CAVs for regularization, effectively mitigating spurious features both locally and globally. To the best of our knowledge, our research is the first to consider the non-linearity of spurious features in model bias regularization. Our results show that CCSR outperforms existing methods on several benchmarks, e.g., Waterbirds, CatDogs, and CelebA-Collars datasets, under the conditions for both with and without group labels on the validation dataset, even when minority samples are absent in the training dataset. These findings highlight the potential of CCSR to improve model robustness and generalization.
format Article
id doaj-art-43c3d65c4b1341f4b5c1969ed37c874f
institution Kabale University
issn 2169-3536
language English
publishDate 2024-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-43c3d65c4b1341f4b5c1969ed37c874f2024-11-22T00:01:02ZengIEEEIEEE Access2169-35362024-01-011217029517030810.1109/ACCESS.2024.347745410713328Debiased Learning via Composed Conceptual Sensitivity RegularizationSunghwan Joo0https://orcid.org/0000-0003-2176-917XTaesup Moon1https://orcid.org/0000-0002-9257-6503Department of Electrical and Computer Engineering, Sungkyunkwan University, Suwon, South KoreaDepartment of Electrical and Computer Engineering, Seoul National University, Seoul, South KoreaDeep neural networks often rely on spurious features, which are attributes correlated with class labels but irrelevant to the actual task, leading to poor generalization when these features are absent. To train classifiers that are not biased towards spurious features, recent research has leveraged explainable AI (XAI) techniques to identify and modify model behavior. Specifically, Concept Activation Vectors (CAVs), which indicate the direction toward specific concepts in the embedding space, were used to measure and regularize the conceptual sensitivity of the classifier, thereby reducing its reliance on spurious features. However, these approaches struggle with non-linear or high-dimensional spurious correlations due to the use of linear CAVs in previous works. In this paper, we propose Composite Conceptual Sensitivity Regularization (CCSR), a novel method designed to address these limitations. CCSR utilizes concept gradients to assign individualized CAVs for each sample, enabling the handling of non-linearly distributed spurious features in embedding space. Additionally, our method employs multiple CAVs for regularization, effectively mitigating spurious features both locally and globally. To the best of our knowledge, our research is the first to consider the non-linearity of spurious features in model bias regularization. Our results show that CCSR outperforms existing methods on several benchmarks, e.g., Waterbirds, CatDogs, and CelebA-Collars datasets, under the conditions for both with and without group labels on the validation dataset, even when minority samples are absent in the training dataset. These findings highlight the potential of CCSR to improve model robustness and generalization.https://ieeexplore.ieee.org/document/10713328/Computer visionexplainable AIconcept activation vectorspurious correlationgroup distributionally robust optimization
spellingShingle Sunghwan Joo
Taesup Moon
Debiased Learning via Composed Conceptual Sensitivity Regularization
IEEE Access
Computer vision
explainable AI
concept activation vector
spurious correlation
group distributionally robust optimization
title Debiased Learning via Composed Conceptual Sensitivity Regularization
title_full Debiased Learning via Composed Conceptual Sensitivity Regularization
title_fullStr Debiased Learning via Composed Conceptual Sensitivity Regularization
title_full_unstemmed Debiased Learning via Composed Conceptual Sensitivity Regularization
title_short Debiased Learning via Composed Conceptual Sensitivity Regularization
title_sort debiased learning via composed conceptual sensitivity regularization
topic Computer vision
explainable AI
concept activation vector
spurious correlation
group distributionally robust optimization
url https://ieeexplore.ieee.org/document/10713328/
work_keys_str_mv AT sunghwanjoo debiasedlearningviacomposedconceptualsensitivityregularization
AT taesupmoon debiasedlearningviacomposedconceptualsensitivityregularization