We present DeepCABAC, a novel context-adaptive binary arithmetic coder for compressing deep neural networks. It quantizes each weight parameter by minimizing a weighted rate-distortion function, which implicitly takes the impact of quantization on to the accuracy of the network into account. Subsequently, it compresses the quantized values into a bitstream representation with minimal redundancies. We show that DeepCABAC is able to reach very high compression ratios across a wide set of different network architectures and datasets. For instance, we are able to compress by x63.6 the VGG16 ImageNet model with no loss of accuracy, thus being able to represent the entire network with merely 8.7MB.
@article{arxiv.1905.08318,
title = {DeepCABAC: Context-adaptive binary arithmetic coding for deep neural network compression},
author = {Simon Wiedemann and Heiner Kirchhoffer and Stefan Matlage and Paul Haase and Arturo Marban and Talmaj Marinc and David Neumann and Ahmed Osman and Detlev Marpe and Heiko Schwarz and Thomas Wiegand and Wojciech Samek},
journal= {arXiv preprint arXiv:1905.08318},
year = {2019}
}
Comments
ICML 2019, Joint Workshop on On-Device Machine Learning and Compact Deep Neural Network Representations (ODML-CDNNR)