OS9-2 Multi-Valued Quantization Neural Networks toward Hardware Implementation

Title:	OS9-2 Multi-Valued Quantization Neural Networks toward Hardware Implementation
Publication:	ICAROB2017
Volume:	22
Pages:	132-135
ISSN:	2188-7829
DOI:	10.5954/ICAROB.2017.OS9-2
Author(s):	Yoshiya Aratani, Yeoh Yoeng Jye, Akihiro Suzuki, Daisuke Shuto, Takashi Morie, Hakaru Tamukoh
Publication Date:	January 19, 2017
Keywords:	Deep Learning, CNN, MVQ, BinaryConnect, hardware, noise
Abstract:	This paper proposes a Multi-Valued Quantization (MVQ) method of connecting weight for efficient hardware implementation of Convolutional Neural Networks (CNNs). The proposed method multiplies an input value by a multi-valued quantized weight during the forward and backward propagations, while retaining precision of the stored weights for the update process. In the both propagation processes, multipliers can be replaced with adders and shifters by setting appropriate quantized weights. We train two- to six-valued quantization CNNs with MNIST and CIFAR10 dataset to compare the performance of them with a 32-bit floating point CNN. In the four-valued quantization, random noise is added to the quantized weight to improve the performance of generalization ability. In addition, the robustness of MVQ CNN to noise is evaluated. Experimental results show that the MVQ CNNs achieve better learning accuracy than the floating point CNN and the four-valued CNN is highly robust to the noise.
PDF File:	https://alife-robotics.co.jp/members2017/icarob/data/html/data/OS_pdf/OS9/OS9-2.pdf
Copyright:	© The authors. This article is distributed under the terms of the Creative Commons Attribution License 4.0, which permits non-commercial use, distribution and reproduction in any medium, provided the original work is properly cited. See for details: https://creativecommons.org/licenses/by-nc/4.0/

HOME