Dream Distillation: A Data-Independent Model Compression Framework
Machine Learning
2019-05-20 v1 Computer Vision and Pattern Recognition
Machine Learning
Abstract
Model compression is eminently suited for deploying deep learning on IoT-devices. However, existing model compression techniques rely on access to the original or some alternate dataset. In this paper, we address the model compression problem when no real data is available, e.g., when data is private. To this end, we propose Dream Distillation, a data-independent model compression framework. Our experiments show that Dream Distillation can achieve 88.5% accuracy on the CIFAR-10 test set without actually training on the original data!
Keywords
Cite
@article{arxiv.1905.07072,
title = {Dream Distillation: A Data-Independent Model Compression Framework},
author = {Kartikeya Bhardwaj and Naveen Suda and Radu Marculescu},
journal= {arXiv preprint arXiv:1905.07072},
year = {2019}
}
Comments
Presented at the ICML 2019 Joint Workshop on On-Device Machine Learning & Compact Deep Neural Network Representations (ODML-CDNNR)