FADRM: Fast and Accurate Data Residual Matching for Dataset Distillation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Cui, Jiacheng, Bi, Xinyue, Luo, Yaxin, Zhao, Xiaohan, Liu, Jiacheng, Shen, Zhiqiang
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912458323525632
author Cui, Jiacheng
Bi, Xinyue
Luo, Yaxin
Zhao, Xiaohan
Liu, Jiacheng
Shen, Zhiqiang
author_facet Cui, Jiacheng
Bi, Xinyue
Luo, Yaxin
Zhao, Xiaohan
Liu, Jiacheng
Shen, Zhiqiang
contents Residual connection has been extensively studied and widely applied at the model architecture level. However, its potential in the more challenging data-centric approaches remains unexplored. In this work, we introduce the concept of Data Residual Matching for the first time, leveraging data-level skip connections to facilitate data generation and mitigate data information vanishing. This approach maintains a balance between newly acquired knowledge through pixel space optimization and existing core local information identification within raw data modalities, specifically for the dataset distillation task. Furthermore, by incorporating optimization-level refinements, our method significantly improves computational efficiency, achieving superior performance while reducing training time and peak GPU memory usage by 50%. Consequently, the proposed method Fast and Accurate Data Residual Matching for Dataset Distillation (FADRM) establishes a new state-of-the-art, demonstrating substantial improvements over existing methods across multiple dataset benchmarks in both efficiency and effectiveness. For instance, with ResNet-18 as the student model and a 0.8% compression ratio on ImageNet-1K, the method achieves 47.7% test accuracy in single-model dataset distillation and 50.0% in multi-model dataset distillation, surpassing RDED by +5.7% and outperforming state-of-the-art multi-model approaches, EDC and CV-DD, by +1.4% and +4.0%. Code is available at: https://github.com/Jiacheng8/FADRM.
format Preprint
id arxiv_https___arxiv_org_abs_2506_24125
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle FADRM: Fast and Accurate Data Residual Matching for Dataset Distillation
Cui, Jiacheng
Bi, Xinyue
Luo, Yaxin
Zhao, Xiaohan
Liu, Jiacheng
Shen, Zhiqiang
Computer Vision and Pattern Recognition
Artificial Intelligence
Residual connection has been extensively studied and widely applied at the model architecture level. However, its potential in the more challenging data-centric approaches remains unexplored. In this work, we introduce the concept of Data Residual Matching for the first time, leveraging data-level skip connections to facilitate data generation and mitigate data information vanishing. This approach maintains a balance between newly acquired knowledge through pixel space optimization and existing core local information identification within raw data modalities, specifically for the dataset distillation task. Furthermore, by incorporating optimization-level refinements, our method significantly improves computational efficiency, achieving superior performance while reducing training time and peak GPU memory usage by 50%. Consequently, the proposed method Fast and Accurate Data Residual Matching for Dataset Distillation (FADRM) establishes a new state-of-the-art, demonstrating substantial improvements over existing methods across multiple dataset benchmarks in both efficiency and effectiveness. For instance, with ResNet-18 as the student model and a 0.8% compression ratio on ImageNet-1K, the method achieves 47.7% test accuracy in single-model dataset distillation and 50.0% in multi-model dataset distillation, surpassing RDED by +5.7% and outperforming state-of-the-art multi-model approaches, EDC and CV-DD, by +1.4% and +4.0%. Code is available at: https://github.com/Jiacheng8/FADRM.
title FADRM: Fast and Accurate Data Residual Matching for Dataset Distillation
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2506.24125