Data-Free Knowledge Distillation with Soft Targeted Transfer Set Synthesis
arXiv:2104.04868
Abstract
Knowledge distillation (KD) has proved to be an effective approach for deep neural network compression, which learns a compact network (student) by transferring the knowledge from a pre-trained, over-parameterized network (teacher). In traditional KD, the transferred knowledge is usually obtained by feeding training samples to the teacher network to obtain the class probabilities. However, the original training dataset is not always available due to storage costs or privacy issues. In this study, we propose a novel data-free KD approach by modeling the intermediate feature space of the teacher with a multivariate normal distribution and leveraging the soft targeted labels generated by the distribution to synthesize pseudo samples as the transfer set. Several student networks trained with these synthesized transfer sets present competitive performance compared to the networks trained with the original training set and other data-free KD approaches.
Accepted to AAAI 2021
References in corpus (9)
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- Distilling the Knowledge in a Neural Network
- Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation
- FitNets: Hints for Thin Deep Nets
- Compressing Deep Convolutional Networks using Vector Quantization
- Pruning Filters for Efficient ConvNets
- Data-Free Knowledge Distillation for Deep Neural Networks
- Variational Information Distillation for Knowledge Transfer
- Uncertainty-Aware Multi-Shot Knowledge Distillation for Image-Based Object Re-Identification