Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Revisiting Knowledge Distillation via Label Smoothing Regularization

About

Knowledge Distillation (KD) aims to distill the knowledge of a cumbersome teacher model into a lightweight student model. Its success is generally attributed to the privileged information on similarities among categories provided by the teacher model, and in this sense, only strong teacher models are deployed to teach weaker students in practice. In this work, we challenge this common belief by following experimental observations: 1) beyond the acknowledgment that the teacher can improve the student, the student can also enhance the teacher significantly by reversing the KD procedure; 2) a poorly-trained teacher with much lower accuracy than the student can still improve the latter significantly. To explain these observations, we provide a theoretical analysis of the relationships between KD and label smoothing regularization. We prove that 1) KD is a type of learned label smoothing regularization and 2) label smoothing regularization provides a virtual teacher model for KD. From these results, we argue that the success of KD is not fully due to the similarity information between categories from teachers, but also to the regularization of soft targets, which is equally or even more important. Based on these analyses, we further propose a novel Teacher-free Knowledge Distillation (Tf-KD) framework, where a student model learns from itself or manuallydesigned regularization distribution. The Tf-KD achieves comparable performance with normal KD from a superior teacher, which is well applied when a stronger teacher model is unavailable. Meanwhile, Tf-KD is generic and can be directly deployed for training deep neural networks. Without any extra computation cost, Tf-KD achieves up to 0.65\% improvement on ImageNet over well-established baseline models, which is superior to label smoothing regularization.

Li Yuan, Francis E.H.Tay, Guilin Li, Tao Wang, Jiashi Feng• 2019

Related benchmarks

TaskDatasetResultRank
Image ClassificationCIFAR-100 (test)
Accuracy77.91
3518
Image ClassificationCIFAR-10 (test)
Accuracy95.05
3381
Object DetectionPASCAL VOC 2007 (test)--
844
Image ClassificationTinyImageNet (test)--
562
Audio ClassificationESC-50
Accuracy75.46
461
Image ClassificationCIFAR-100 (test)--
429
Image ClassificationILSVRC 2012 (val)
Top-1 Accuracy70.52
156
Image ClassificationImageNet (val)
Top-1 Error0.22
46
Acoustic ClassificationDeepShip
Accuracy60.13
25
Image ClassificationCIFAR-100 (test)
Top-1 Acc82.67
24
Showing 10 of 13 rows

Other info

Follow for update