Scholarlink

<p>Dataset lightweighting is the process of constructing a small dataset from a large original dataset to reduce the pressure of data storage, distribution, and model training. Dataset distillation(DD) is an emerging dataset lightweight method in recent years. This paper mainly focuses on DD problem. The mainstream algorithms of DD can be divided into several categories: meta learning framework, parameter matching framework, distribution matching framework and generative model-based DD. In this paper, we provide a comprehensive review of the distillation methods of datasets since 2018 in the above order. At the end of the article, we present some discussions on the current state of development in this field and suggest some potential future research directions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Dataset Distillation: Recent Advances of Methods and Challenges

  • 期刊论文

  • 发表日期：

    2025年10月1日

查看全文
  • Muyang Li,
  • Yong Shi

摘要

Dataset lightweighting is the process of constructing a small dataset from a large original dataset to reduce the pressure of data storage, distribution, and model training. Dataset distillation(DD) is an emerging dataset lightweight method in recent years. This paper mainly focuses on DD problem. The mainstream algorithms of DD can be divided into several categories: meta learning framework, parameter matching framework, distribution matching framework and generative model-based DD. In this paper, we provide a comprehensive review of the distillation methods of datasets since 2018 in the above order. At the end of the article, we present some discussions on the current state of development in this field and suggest some potential future research directions.