Skip to content

Latest commit

 

History

History

wsld

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 

WSLD

Rethinking Soft Labels for Knowledge Distillation: A Bias-Variance Tradeoff Perspective

Abstract

Knowledge distillation is an effective approach to leverage a well-trained network or an ensemble of them, named as the teacher, to guide the training of a student network. The outputs from the teacher network are used as soft labels for supervising the training of a new network. Recent studies (Muller et al., 2019; Yuan ¨ et al., 2020) revealed an intriguing property of the soft labels that making labels soft serves as a good regularization to the student network. From the perspective of statistical learning, regularization aims to reduce the variance, however how bias and variance change is not clear for training with soft labels. In this paper, we investigate the bias-variance tradeoff brought by distillation with soft labels. Specifically, we observe that during training the bias-variance tradeoff varies sample-wisely. Further, under the same distillation temperature setting, we observe that the distillation performance is negatively associated with the number of some specific samples, which are named as regularization samples since these samples lead to bias increasing and variance decreasing. Nevertheless, we empirically find that completely filtering out regularization samples also deteriorates distillation performance. Our discoveries inspired us to propose the novel weighted soft labels to help the network adaptively handle the sample-wise biasvariance tradeoff. Experiments on standard evaluation benchmarks validate the effectiveness of our method.

pipeline

Results and models

Classification

Location Dataset Teacher Student Acc Acc(T) Acc(S) Config Download
cls head ImageNet resnet34 resnet18 71.54 73.62 69.90 config teacher |model | log

Citation

@inproceedings{zhou2021wsl,
  title={Rethinking soft labels for knowledge distillation: a bias-variance tradeoff perspective},
  author={Helong, Zhou and Liangchen, Song and Jiajie, Chen and Ye, Zhou and Guoli, Wang and Junsong, Yuan and Qian Zhang},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year={2021}
}