Jump-teaching: Combating Sample Selection Bias via Temporal Disagreement

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ji, Kangye, Cheng, Fei, Wang, Zeqing, Zhang, Qichang, Huang, Bohu
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915730282250240
author Ji, Kangye
Cheng, Fei
Wang, Zeqing
Zhang, Qichang
Huang, Bohu
author_facet Ji, Kangye
Cheng, Fei
Wang, Zeqing
Zhang, Qichang
Huang, Bohu
contents Sample selection is a straightforward technique to combat noisy labels, aiming to prevent mislabeled samples from degrading the robustness of neural networks. However, existing methods mitigate compounding selection bias either by leveraging dual-network disagreement or additional forward propagations, leading to multiplied training overhead. To address this challenge, we introduce $\textit{Jump-teaching}$, an efficient sample selection framework for debiased model update and simplified selection criterion. Based on a key observation that a neural network exhibits significant disagreement across different training iterations, Jump-teaching proposes a jump-manner model update strategy to enable self-correction of selection bias by harnessing temporal disagreement, eliminating the need for multi-network or multi-round training. Furthermore, we employ a sample-wise selection criterion building on the intra variance of a decomposed single loss for a fine-grained selection without relying on batch-wise ranking or dataset-wise modeling. Extensive experiments demonstrate that Jump-teaching outperforms state-of-the-art counterparts while achieving a nearly overhead-free selection procedure, which boosts training speed by up to $4.47\times$ and reduces peak memory footprint by $54\%$.
format Preprint
id arxiv_https___arxiv_org_abs_2405_17137
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Jump-teaching: Combating Sample Selection Bias via Temporal Disagreement
Ji, Kangye
Cheng, Fei
Wang, Zeqing
Zhang, Qichang
Huang, Bohu
Computer Vision and Pattern Recognition
Sample selection is a straightforward technique to combat noisy labels, aiming to prevent mislabeled samples from degrading the robustness of neural networks. However, existing methods mitigate compounding selection bias either by leveraging dual-network disagreement or additional forward propagations, leading to multiplied training overhead. To address this challenge, we introduce $\textit{Jump-teaching}$, an efficient sample selection framework for debiased model update and simplified selection criterion. Based on a key observation that a neural network exhibits significant disagreement across different training iterations, Jump-teaching proposes a jump-manner model update strategy to enable self-correction of selection bias by harnessing temporal disagreement, eliminating the need for multi-network or multi-round training. Furthermore, we employ a sample-wise selection criterion building on the intra variance of a decomposed single loss for a fine-grained selection without relying on batch-wise ranking or dataset-wise modeling. Extensive experiments demonstrate that Jump-teaching outperforms state-of-the-art counterparts while achieving a nearly overhead-free selection procedure, which boosts training speed by up to $4.47\times$ and reduces peak memory footprint by $54\%$.
title Jump-teaching: Combating Sample Selection Bias via Temporal Disagreement
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2405.17137