Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DeepCore: A Comprehensive Library for Coreset Selection in Deep Learning

Chengcheng Guo, Bo Zhao, Yanbing Bai

TL;DR

This work tackles coreset selection in deep learning by introducing DeepCore, a modular PyTorch library that re-implements a wide range of coreset methods under a unified framework. It provides a rigorous, fair benchmark of methods across CIFAR10 and ImageNet, revealing that random selection remains a strong baseline and that method performance is highly sensitive to training settings, coreset size, and architecture. The paper categorizes core methodologies, analyzes cross-architecture generalization, and examines sensitivity to pre-trained models, delivering practical insights for fair comparison and reproducibility in the field. Overall, DeepCore enables standardized evaluation of data-efficient training strategies, guiding future research toward robust, generalizable coreset selection techniques.

Abstract

Coreset selection, which aims to select a subset of the most informative training samples, is a long-standing learning problem that can benefit many downstream tasks such as data-efficient learning, continual learning, neural architecture search, active learning, etc. However, many existing coreset selection methods are not designed for deep learning, which may have high complexity and poor generalization performance. In addition, the recently proposed methods are evaluated on models, datasets, and settings of different complexities. To advance the research of coreset selection in deep learning, we contribute a comprehensive code library, namely DeepCore, and provide an empirical study on popular coreset selection methods on CIFAR10 and ImageNet datasets. Extensive experiments on CIFAR10 and ImageNet datasets verify that, although various methods have advantages in certain experiment settings, random selection is still a strong baseline.

DeepCore: A Comprehensive Library for Coreset Selection in Deep Learning

TL;DR

This work tackles coreset selection in deep learning by introducing DeepCore, a modular PyTorch library that re-implements a wide range of coreset methods under a unified framework. It provides a rigorous, fair benchmark of methods across CIFAR10 and ImageNet, revealing that random selection remains a strong baseline and that method performance is highly sensitive to training settings, coreset size, and architecture. The paper categorizes core methodologies, analyzes cross-architecture generalization, and examines sensitivity to pre-trained models, delivering practical insights for fair comparison and reproducibility in the field. Overall, DeepCore enables standardized evaluation of data-efficient training strategies, guiding future research toward robust, generalizable coreset selection techniques.

Abstract

Coreset selection, which aims to select a subset of the most informative training samples, is a long-standing learning problem that can benefit many downstream tasks such as data-efficient learning, continual learning, neural architecture search, active learning, etc. However, many existing coreset selection methods are not designed for deep learning, which may have high complexity and poor generalization performance. In addition, the recently proposed methods are evaluated on models, datasets, and settings of different complexities. To advance the research of coreset selection in deep learning, we contribute a comprehensive code library, namely DeepCore, and provide an empirical study on popular coreset selection methods on CIFAR10 and ImageNet datasets. Extensive experiments on CIFAR10 and ImageNet datasets verify that, although various methods have advantages in certain experiment settings, random selection is still a strong baseline.
Paper Structure (27 sections, 9 equations, 1 figure, 4 tables)

This paper contains 27 sections, 9 equations, 1 figure, 4 tables.

Table of Contents

  1. Introduction
  2. Review of Coreset Selection Methods
  3. Problem Statement
  4. Survey: Methodologies
  5. Geometry Based Methods
  6. Uncertainty Based Methods
  7. Error/Loss Based Methods
  8. Decision Boundary Based Methods
  9. Gradient Matching Based Methods
  10. Bilevel Optimization Based Methods
  11. Submodularity Based Methods
  12. Proxy Based Methods
  13. Survey: Applications
  14. Data-efficient Learning.
  15. Continual Learning.
  16. ...and 12 more sections

Figures (1)

  • Figure 1: Coreset selection performances in curves on CIFAR10. We train randomly initialized ResNet-18 on the coresets of CIFAR10 produced by different methods and then test on the real testing set. Detailed numbers are provided in Tab. \ref{['tab:cifar']}.