Feature Clipping for Uncertainty Calibration

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Tao, Linwei, Dong, Minjing, Xu, Chang
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913564678160384
author Tao, Linwei
Dong, Minjing
Xu, Chang
author_facet Tao, Linwei
Dong, Minjing
Xu, Chang
contents Deep neural networks (DNNs) have achieved significant success across various tasks, but ensuring reliable uncertainty estimates, known as model calibration, is crucial for their safe and effective deployment. Modern DNNs often suffer from overconfidence, leading to miscalibration. We propose a novel post-hoc calibration method called feature clipping (FC) to address this issue. FC involves clipping feature values to a specified threshold, effectively increasing entropy in high calibration error samples while maintaining the information in low calibration error samples. This process reduces the overconfidence in predictions, improving the overall calibration of the model. Our extensive experiments on datasets such as CIFAR-10, CIFAR-100, and ImageNet, and models including CNNs and transformers, demonstrate that FC consistently enhances calibration performance. Additionally, we provide a theoretical analysis that validates the effectiveness of our method. As the first calibration technique based on feature modification, feature clipping offers a novel approach to improving model calibration, showing significant improvements over both post-hoc and train-time calibration methods and pioneering a new avenue for feature-based model calibration.
format Preprint
id arxiv_https___arxiv_org_abs_2410_19796
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Feature Clipping for Uncertainty Calibration
Tao, Linwei
Dong, Minjing
Xu, Chang
Computer Vision and Pattern Recognition
Machine Learning
Deep neural networks (DNNs) have achieved significant success across various tasks, but ensuring reliable uncertainty estimates, known as model calibration, is crucial for their safe and effective deployment. Modern DNNs often suffer from overconfidence, leading to miscalibration. We propose a novel post-hoc calibration method called feature clipping (FC) to address this issue. FC involves clipping feature values to a specified threshold, effectively increasing entropy in high calibration error samples while maintaining the information in low calibration error samples. This process reduces the overconfidence in predictions, improving the overall calibration of the model. Our extensive experiments on datasets such as CIFAR-10, CIFAR-100, and ImageNet, and models including CNNs and transformers, demonstrate that FC consistently enhances calibration performance. Additionally, we provide a theoretical analysis that validates the effectiveness of our method. As the first calibration technique based on feature modification, feature clipping offers a novel approach to improving model calibration, showing significant improvements over both post-hoc and train-time calibration methods and pioneering a new avenue for feature-based model calibration.
title Feature Clipping for Uncertainty Calibration
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2410.19796