.png){kind=icon}
.png){kind=icon}
Return
Conditional pseudo-supervised contrast for data-Free knowledge distillation
.png){kind=icon}
PRE
AI
DOI：10.1016/j.patcog.2023.109781.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Data-free knowledge distillation (DFKD) is an effective manner to solve model compression and trans-mission restrictions while retaining privacy protection, which has attracted extensive attention in recent years. Currently, the majority of existing methods utilize a generator to synthesize images to support the distillation. Although the current methods have achieved great success, there are still many issues to be explored. Firstly, the outstanding performance of supervised learning in deep learning drives us to explore a pseudo-supervised paradigm on DFKD. Secondly, current synthesized methods cannot distin-guish the distributions of different categories of samples, thus producing ambiguous samples that may lead to an incorrect evaluation by the teacher. Besides, current methods cannot optimize the category -wise diversity samples, which will hinder the student model learning from diverse samples and further achieving better performance. In this paper, to address the above limitations, we propose a novel learning paradigm, i.e., conditional pseudo-supervised contrast for data-free knowledge distillation (CPSC-DFKD). The primary innovations of CPSC-DFKD are: (1) introducing a conditional generative adversarial network to synthesize category-specific diverse images for pseudo-supervised learning, (2) improving the mod-ules of the generator to distinguish the distributions of different categories, and (3) proposing pseudo -supervised contrastive learning based on teacher and student views to enhance diversity. Comprehensive experiments on three commonly-used datasets validate the performance lift of both the student and gen-erator brought by CPSC-DFKD. The code is available at https://github.com/RoryShao/CPSC-DFKD.git & COPY; 2023 Elsevier Ltd. All rights reserved.
Keywords:
Model compression
Knowledge distillation
Representation learning
Contrastive learning
Privacy protection
Journal
IF:
7.6
Papers:
1.3W
Citations:
4.5W