Abstract:Data-free knowledge distillation is able to utilize the knowledge learned by a large teacher network to augment the training of a smaller student network without accessing the original training data, avoiding privacy, security, and proprietary risks in real applications. In this line of research, existing methods typically follow an inversion-and-distillation paradigm in which a generative adversarial network on-the-fly trained with the guidance of the pre-trained teacher network is used to synthesize a large-scale sample set for knowledge distillation. In this paper, we reexamine this common data-free knowledge distillation paradigm, showing that there is considerable room to improve the overall training efficiency through a lens of ``small-scale inverted data for knowledge distillation". In light of three empirical observations indicating the importance of how to balance class distributions in terms of synthetic sample diversity and difficulty during both data inversion and distillation processes, we propose Small Scale Data-free Knowledge Distillation SSD-KD. In formulation, SSD-KD introduces a modulating function to balance synthetic samples and a priority sampling function to select proper samples, facilitated by a dynamic replay buffer and a reinforcement learning strategy. As a result, SSD-KD can perform distillation training conditioned on an extremely small scale of synthetic samples (e.g., 10X less than the original training data scale), making the overall training efficiency one or two orders of magnitude faster than many mainstream methods while retaining superior or competitive model performance, as demonstrated on popular image classification and semantic segmentation benchmarks. The code is available at <a class="link-external link-https" href="https://github.com/OSVAI/SSD-KD" rel="external noopener nofollow">this https URL</a>.

Attention Based Data Augmentation for Knowledge Distillation with Few Data

Knowledge Distillation Method for Surface Defect Detection.

Research on Knowledge Distillation Algorithm of Object Detection

Role-Wise Data Augmentation for Knowledge Distillation

Understanding the Effect of Data Augmentation on Knowledge Distillation

SAKD: Sparse attention knowledge distillation

Why does Knowledge Distillation Work? Rethink its Attention and Fidelity Mechanism

Comparative Knowledge Distillation

An Empirical Analysis of the Impact of Data Augmentation on Knowledge Distillation

Data-Free Adversarial Distillation

Towards Effective Data-Free Knowledge Distillation via Diverse Diffusion Augmentation

Knowledge Augmentation for Distillation: A General and Effective Approach to Enhance Knowledge Distillation

Small Scale Data-Free Knowledge Distillation

Learning to Augment for Data-scarce Domain BERT Knowledge Distillation

Teaching What You Should Teach: A Data-Based Distillation Method

Attention and feature transfer based knowledge distillation

Isotonic Data Augmentation for Knowledge Distillation

Improve Knowledge Distillation via Label Revision and Data Selection

A-A KD: Attention and Activation Knowledge Distillation

Show, Attend and Distill:Knowledge Distillation via Attention-based Feature Matching