Rethinking Soft Labels for Knowledge Distillation: A Bias-Variance Tradeoff Perspective
arXiv:2102.00650
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 \citep{muller2019does,yuan2020revisiting} 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 bias-variance tradeoff. Experiments on standard evaluation benchmarks validate the effectiveness of our method. Our code is available at \url{https://github.com/bellymonster/Weighted-Soft-Label-Distillation}.
arXiv commentICLR 2021
References in corpus (13)
- Distilling the Knowledge in a Neural Network
- Reconciling modern machine learning practice and the bias-variance trade-off
- Do Deep Nets Really Need to be Deep?
- Regularizing Neural Networks by Penalizing Confident Output Distributions
- Like What You Like: Knowledge Distill via Neuron Selectivity Transfer
- Bayesian Dark Knowledge
- Towards Understanding Knowledge Distillation
- Uniform convergence may be unable to explain generalization in deep learning
- Rethinking Bias-Variance Trade-off for Generalization of Neural Networks
- A Modern Take on the Bias-Variance Tradeoff in Neural Networks
- Knowledge Consistency between Neural Networks and Beyond
- Preparing Lessons: Improve Knowledge Distillation with Better Supervision
- Learning Efficient Detector with Semi-supervised Adaptive Distillation