TY - GEN
T1 - Reducing the model order of deep neural networks using information theory
AU - Tu, Ming
AU - Berisha, Visar
AU - Cao, Yu
AU - Seo, Jae-sun
N1 - Funding Information:
This research was supported in part by the Office of Naval Research grant N000141410722 (Berisha), an ASU-Mayo seed grant, and a hardware grant from NVIDIA.
Publisher Copyright:
© 2016 IEEE.
PY - 2016/9/2
Y1 - 2016/9/2
N2 - Deep neural networks are typically represented by a much larger number of parameters than shallow models, making them prohibitive for small footprint devices. Recent research shows that there is considerable redundancy in the parameter space of deep neural networks. In this paper, we propose a method to compress deep neural networks by using the Fisher Information metric, which we estimate through a stochastic optimization method that keeps track of second-order information in the network. We first remove unimportant parameters and then use non-uniform fixed point quantization to assign more bits to parameters with higher Fisher Information estimates. We evaluate our method on a classification task with a convolutional neural network trained on the MNIST data set. Experimental results show that our method outperforms existing methods for both network pruning and quantization.
AB - Deep neural networks are typically represented by a much larger number of parameters than shallow models, making them prohibitive for small footprint devices. Recent research shows that there is considerable redundancy in the parameter space of deep neural networks. In this paper, we propose a method to compress deep neural networks by using the Fisher Information metric, which we estimate through a stochastic optimization method that keeps track of second-order information in the network. We first remove unimportant parameters and then use non-uniform fixed point quantization to assign more bits to parameters with higher Fisher Information estimates. We evaluate our method on a classification task with a convolutional neural network trained on the MNIST data set. Experimental results show that our method outperforms existing methods for both network pruning and quantization.
UR - http://www.scopus.com/inward/record.url?scp=84988929256&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84988929256&partnerID=8YFLogxK
U2 - 10.1109/ISVLSI.2016.117
DO - 10.1109/ISVLSI.2016.117
M3 - Conference contribution
AN - SCOPUS:84988929256
T3 - Proceedings of IEEE Computer Society Annual Symposium on VLSI, ISVLSI
SP - 93
EP - 98
BT - Proceedings - IEEE Computer Society Annual Symposium on VLSI, ISVLSI 2016
PB - IEEE Computer Society
T2 - 15th IEEE Computer Society Annual Symposium on VLSI, ISVLSI 2016
Y2 - 11 July 2016 through 13 July 2016
ER -