Guardado en:
Detalles Bibliográficos
Autores principales: Zhang, Songming, Luo, Yuxiao, Lyu, Ziyu, Chen, Xiaofeng
Formato: Preprint
Publicado: 2023
Materias:
Acceso en línea:https://arxiv.org/abs/2312.16242
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866912491103059968
author Zhang, Songming
Luo, Yuxiao
Lyu, Ziyu
Chen, Xiaofeng
author_facet Zhang, Songming
Luo, Yuxiao
Lyu, Ziyu
Chen, Xiaofeng
contents Knowledge Distillation (KD) transfers knowledge from large models to small models and has recently achieved remarkable success. However, the reliability of existing KD methods in real-world applications, especially under distribution shift, remains underexplored. Distribution shift refers to the data distribution drifts between the training and testing phases, and this can adversely affect the efficacy of KD. In this paper, we propose a unified and systematic framework \textsc{ShiftKD} to benchmark KD against two general distributional shifts: diversity and correlation shift. The evaluation benchmark covers more than 30 methods from algorithmic, data-driven, and optimization perspectives for five benchmark datasets. Our development of \textsc{ShiftKD} conducts extensive experiments and reveals strengths and limitations of current SOTA KD methods. More importantly, we thoroughly analyze key factors in student model training process, including data augmentation, pruning methods, optimizers, and evaluation metrics. We believe \textsc{ShiftKD} could serve as an effective benchmark for assessing KD in real-world scenarios, thus driving the development of more robust KD methods in response to evolving demands. The code will be made available upon publication.
format Preprint
id arxiv_https___arxiv_org_abs_2312_16242
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle ShiftKD: Benchmarking Knowledge Distillation under Distribution Shift
Zhang, Songming
Luo, Yuxiao
Lyu, Ziyu
Chen, Xiaofeng
Machine Learning
Knowledge Distillation (KD) transfers knowledge from large models to small models and has recently achieved remarkable success. However, the reliability of existing KD methods in real-world applications, especially under distribution shift, remains underexplored. Distribution shift refers to the data distribution drifts between the training and testing phases, and this can adversely affect the efficacy of KD. In this paper, we propose a unified and systematic framework \textsc{ShiftKD} to benchmark KD against two general distributional shifts: diversity and correlation shift. The evaluation benchmark covers more than 30 methods from algorithmic, data-driven, and optimization perspectives for five benchmark datasets. Our development of \textsc{ShiftKD} conducts extensive experiments and reveals strengths and limitations of current SOTA KD methods. More importantly, we thoroughly analyze key factors in student model training process, including data augmentation, pruning methods, optimizers, and evaluation metrics. We believe \textsc{ShiftKD} could serve as an effective benchmark for assessing KD in real-world scenarios, thus driving the development of more robust KD methods in response to evolving demands. The code will be made available upon publication.
title ShiftKD: Benchmarking Knowledge Distillation under Distribution Shift
topic Machine Learning
url https://arxiv.org/abs/2312.16242