Guardado en:
Detalles Bibliográficos
Autores principales: Zhong, Xinhao, Chen, Bin, Fang, Hao, Gu, Xulin, Xia, Shu-Tao, Yang, En-Hui
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:https://arxiv.org/abs/2412.09945
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866908368060284928
author Zhong, Xinhao
Chen, Bin
Fang, Hao
Gu, Xulin
Xia, Shu-Tao
Yang, En-Hui
author_facet Zhong, Xinhao
Chen, Bin
Fang, Hao
Gu, Xulin
Xia, Shu-Tao
Yang, En-Hui
contents Dataset distillation (DD) aims to minimize the time and memory consumption needed for training deep neural networks on large datasets, by creating a smaller synthetic dataset that has similar performance to that of the full real dataset. However, current dataset distillation methods often result in synthetic datasets that are excessively difficult for networks to learn from, due to the compression of a substantial amount of information from the original data through metrics measuring feature similarity, e,g., distribution matching (DM). In this work, we introduce conditional mutual information (CMI) to assess the class-aware complexity of a dataset and propose a novel method by minimizing CMI. Specifically, we minimize the distillation loss while constraining the class-aware complexity of the synthetic dataset by minimizing its empirical CMI from the feature space of pre-trained networks, simultaneously. Conducting on a thorough set of experiments, we show that our method can serve as a general regularization method to existing DD methods and improve the performance and training efficiency.
format Preprint
id arxiv_https___arxiv_org_abs_2412_09945
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Going Beyond Feature Similarity: Effective Dataset Distillation based on Class-Aware Conditional Mutual Information
Zhong, Xinhao
Chen, Bin
Fang, Hao
Gu, Xulin
Xia, Shu-Tao
Yang, En-Hui
Computer Vision and Pattern Recognition
Dataset distillation (DD) aims to minimize the time and memory consumption needed for training deep neural networks on large datasets, by creating a smaller synthetic dataset that has similar performance to that of the full real dataset. However, current dataset distillation methods often result in synthetic datasets that are excessively difficult for networks to learn from, due to the compression of a substantial amount of information from the original data through metrics measuring feature similarity, e,g., distribution matching (DM). In this work, we introduce conditional mutual information (CMI) to assess the class-aware complexity of a dataset and propose a novel method by minimizing CMI. Specifically, we minimize the distillation loss while constraining the class-aware complexity of the synthetic dataset by minimizing its empirical CMI from the feature space of pre-trained networks, simultaneously. Conducting on a thorough set of experiments, we show that our method can serve as a general regularization method to existing DD methods and improve the performance and training efficiency.
title Going Beyond Feature Similarity: Effective Dataset Distillation based on Class-Aware Conditional Mutual Information
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2412.09945