Mitigating Spurious Correlations with Causal Logit Perturbation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhou, Xiaoling, Ye, Wei, Xie, Rui, Zhang, Shikun
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908373028438016
author Zhou, Xiaoling
Ye, Wei
Xie, Rui
Zhang, Shikun
author_facet Zhou, Xiaoling
Ye, Wei
Xie, Rui
Zhang, Shikun
contents Deep learning has seen widespread success in various domains such as science, industry, and society. However, it is acknowledged that certain approaches suffer from non-robustness, relying on spurious correlations for predictions. Addressing these limitations is of paramount importance, necessitating the development of methods that can disentangle spurious correlations. {This study attempts to implement causal models via logit perturbations and introduces a novel Causal Logit Perturbation (CLP) framework to train classifiers with generated causal logit perturbations for individual samples, thereby mitigating the spurious associations between non-causal attributes (i.e., image backgrounds) and classes.} {Our framework employs a} perturbation network to generate sample-wise logit perturbations using a series of training characteristics of samples as inputs. The whole framework is optimized by an online meta-learning-based learning algorithm and leverages human causal knowledge by augmenting metadata in both counterfactual and factual manners. Empirical evaluations on four typical biased learning scenarios, including long-tail learning, noisy label learning, generalized long-tail learning, and subpopulation shift learning, demonstrate that CLP consistently achieves state-of-the-art performance. Moreover, visualization results support the effectiveness of the generated causal perturbations in redirecting model attention towards causal image attributes and dismantling spurious associations.
format Preprint
id arxiv_https___arxiv_org_abs_2505_15246
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Mitigating Spurious Correlations with Causal Logit Perturbation
Zhou, Xiaoling
Ye, Wei
Xie, Rui
Zhang, Shikun
Machine Learning
K.3.2
F.4.1
Deep learning has seen widespread success in various domains such as science, industry, and society. However, it is acknowledged that certain approaches suffer from non-robustness, relying on spurious correlations for predictions. Addressing these limitations is of paramount importance, necessitating the development of methods that can disentangle spurious correlations. {This study attempts to implement causal models via logit perturbations and introduces a novel Causal Logit Perturbation (CLP) framework to train classifiers with generated causal logit perturbations for individual samples, thereby mitigating the spurious associations between non-causal attributes (i.e., image backgrounds) and classes.} {Our framework employs a} perturbation network to generate sample-wise logit perturbations using a series of training characteristics of samples as inputs. The whole framework is optimized by an online meta-learning-based learning algorithm and leverages human causal knowledge by augmenting metadata in both counterfactual and factual manners. Empirical evaluations on four typical biased learning scenarios, including long-tail learning, noisy label learning, generalized long-tail learning, and subpopulation shift learning, demonstrate that CLP consistently achieves state-of-the-art performance. Moreover, visualization results support the effectiveness of the generated causal perturbations in redirecting model attention towards causal image attributes and dismantling spurious associations.
title Mitigating Spurious Correlations with Causal Logit Perturbation
topic Machine Learning
K.3.2
F.4.1
url https://arxiv.org/abs/2505.15246