Quantifying the Knowledge in a DNN to Explain Knowledge Distillation for Classification

Compared to traditional learning from scratch, knowledge distillation sometimes makes the DNN achieve superior performance. In this paper, we provide a new perspective to explain the success of knowledge distillation based on the information theory, i.e., quantifying knowledge points encoded in inte...

Ausführliche Beschreibung

Bibliographische Detailangaben
Veröffentlicht in:IEEE transactions on pattern analysis and machine intelligence. - 1979. - 45(2023), 4 vom: 22. Apr., Seite 5099-5113
1. Verfasser: Zhang, Quanshi (VerfasserIn)
Weitere Verfasser: Cheng, Xu, Chen, Yilan, Rao, Zhefan
Format: Online-Aufsatz
Sprache:English
Veröffentlicht: 2023
Zugriff auf das übergeordnete Werk:IEEE transactions on pattern analysis and machine intelligence
Schlagworte:Journal Article