English

Zero-Shot Knowledge Distillation in Deep Networks

Machine Learning 2019-05-21 v1 Computer Vision and Pattern Recognition Machine Learning

Abstract

Knowledge distillation deals with the problem of training a smaller model (Student) from a high capacity source model (Teacher) so as to retain most of its performance. Existing approaches use either the training data or meta-data extracted from it in order to train the Student. However, accessing the dataset on which the Teacher has been trained may not always be feasible if the dataset is very large or it poses privacy or safety concerns (e.g., bio-metric or medical data). Hence, in this paper, we propose a novel data-free method to train the Student from the Teacher. Without even using any meta-data, we synthesize the Data Impressions from the complex Teacher model and utilize these as surrogates for the original training data samples to transfer its learning to Student via knowledge distillation. We, therefore, dub our method "Zero-Shot Knowledge Distillation" and demonstrate that our framework results in competitive generalization performance as achieved by distillation using the actual training data samples on multiple benchmark datasets.

Keywords

Cite

@article{arxiv.1905.08114,
  title  = {Zero-Shot Knowledge Distillation in Deep Networks},
  author = {Gaurav Kumar Nayak and Konda Reddy Mopuri and Vaisakh Shaj and R. Venkatesh Babu and Anirban Chakraborty},
  journal= {arXiv preprint arXiv:1905.08114},
  year   = {2019}
}

Comments

Accepted in ICML 2019, codes will be available at https://github.com/vcl-iisc/ZSKD

R2 v1 2026-06-23T09:13:24.306Z