Enregistré dans:
Détails bibliographiques
Auteurs principaux: Deng, Qiyan, Zheng, Changqian, Qiao, Lianpeng, Wang, Yuping, Chai, Chengliang, Cao, Lei
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2510.27253
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866915588463394816
author Deng, Qiyan
Zheng, Changqian
Qiao, Lianpeng
Wang, Yuping
Chai, Chengliang
Cao, Lei
author_facet Deng, Qiyan
Zheng, Changqian
Qiao, Lianpeng
Wang, Yuping
Chai, Chengliang
Cao, Lei
contents Dataset distillation condenses large datasets into synthetic subsets, achieving performance comparable to training on the full dataset while substantially reducing storage and computation costs. Most existing dataset distillation methods assume that all real instances contribute equally to the process. In practice, real-world datasets contain both informative and redundant or even harmful instances, and directly distilling the full dataset without considering data quality can degrade model performance. In this work, we present Influence-Weighted Distillation IWD, a principled framework that leverages influence functions to explicitly account for data quality in the distillation process. IWD assigns adaptive weights to each instance based on its estimated impact on the distillation objective, prioritizing beneficial data while downweighting less useful or harmful ones. Owing to its modular design, IWD can be seamlessly integrated into diverse dataset distillation frameworks. Our empirical results suggest that integrating IWD tends to improve the quality of distilled datasets and enhance model performance, with accuracy gains of up to 7.8%.
format Preprint
id arxiv_https___arxiv_org_abs_2510_27253
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Not All Instances Are Equally Valuable: Towards Influence-Weighted Dataset Distillation
Deng, Qiyan
Zheng, Changqian
Qiao, Lianpeng
Wang, Yuping
Chai, Chengliang
Cao, Lei
Machine Learning
Artificial Intelligence
Dataset distillation condenses large datasets into synthetic subsets, achieving performance comparable to training on the full dataset while substantially reducing storage and computation costs. Most existing dataset distillation methods assume that all real instances contribute equally to the process. In practice, real-world datasets contain both informative and redundant or even harmful instances, and directly distilling the full dataset without considering data quality can degrade model performance. In this work, we present Influence-Weighted Distillation IWD, a principled framework that leverages influence functions to explicitly account for data quality in the distillation process. IWD assigns adaptive weights to each instance based on its estimated impact on the distillation objective, prioritizing beneficial data while downweighting less useful or harmful ones. Owing to its modular design, IWD can be seamlessly integrated into diverse dataset distillation frameworks. Our empirical results suggest that integrating IWD tends to improve the quality of distilled datasets and enhance model performance, with accuracy gains of up to 7.8%.
title Not All Instances Are Equally Valuable: Towards Influence-Weighted Dataset Distillation
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2510.27253