A Kernel Decomposition Architecture for Binary-weight Convolutional Neural Networks
Authors:
Open Access
- 18 June 2017
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 1-6
Abstract
The binary-weight CNN is one of the most efficient solutions for mobile CNNs. However, a large number of operations are required to process each image. To reduce such a huge operation count, we propose an energy-efficient kernel decomposition architecture, based on the observation that a large number of operations are redundant. In this scheme, all kernels are decomposed into sub-kernels to expose the common parts. By skipping the redundant computations, the operation count for each image was consequently reduced by 47.7%. Furthermore, a low cost bit-width quantization technique was implemented by exploiting the relative scales of the feature data. Experimental results showed that the proposed architecture achieves a 22% energy reduction.Keywords
Funding Information
- National Research Foundation of Korea (No. NRF-2014R1A2A1A05004316, No. 2010-0028680)
This publication has 5 references indexed in Scilit:
- XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks2016
- Cnvlutin: Ineffectual-Neuron-Free Deep Neural Network Computing2016
- EIE: Efficient Inference Engine on Compressed Deep Neural Network2016
- CNN Features Off-the-Shelf: An Astounding Baseline for Recognition2014
- Gradient-based learning applied to document recognitionProceedings of the IEEE, 1998