Presentation | 2017-07-27 Consideration of All Binarized Convolutional Neural Network Masayuki Shimoda, Tomoya Fujii, Haruyoshi Yonekawa, Shimpei Sato, Hiroki Nakahara, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | A pre-trained convolutional neural network (CNN) is a feed-forward computation perspective, which is widely used for the embedded systems, requires high power-and-area efficiency. This paper realizes a binarized CNN which treats only binary values (+1/-1) for the inputs, the weights and the activation value. In this case, the multiplier is replaced into an XNOR circuit instead of a dedicated DSP block. Both inputs and weights are more suitable for hardware implementation. However, first convolutional layer still calculates in integer precision, since input value is not binarized one. In this paper, we transform input value into maps of which each pixel is 1 bit precision. The proposed method enables a binarized CNN to use bitwise operation in all layers of convolution. We call this all binarized CNN. We conduct experiment on comparing all binarized CNN, floating-point CNN and binarized CNN. Since all binarized CNN do not need dedicated DSP block, area of all binarized CNN is smaller than that of conventional binarized CNN. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Neural Network / Convolutional Neural Network / Binary Convolutional Neural Network |
Paper # | CPSY2017-28 |
Date of Issue | 2017-07-19 (CPSY) |
Conference Information | |
Committee | CPSY / DC / IPSJ-ARC |
---|---|
Conference Date | 2017/7/26(3days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Akita Atorion-Building (Akita) |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Parallel, Distributed and Cooperative Processing |
Chair | Koji Nakano(Hiroshima Univ.) / Michiko Inoue(NAIST) |
Vice Chair | Hidetsugu Irie(Univ. of Tokyo) / Takashi Miyoshi(Fujitsu) / Satoshi Fukumoto(Tokyo Metropolitan Univ.) |
Secretary | Hidetsugu Irie(Utsunomiya Univ.) / Takashi Miyoshi(Hokkaido Univ.) / Satoshi Fukumoto(Kyoto Sangyo Univ.) / (Tokyo Inst. of Tech.) |
Assistant | Yasuaki Ito(Hiroshima Univ.) / Tomoaki Tsumura(Nagoya Inst. of Tech.) / Masayuki Arai(Nihon Univ.) |
Paper Information | |
Registration To | Technical Committee on Computer Systems / Technical Committee on Dependable Computing / Special Interest Group on System Architecture |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Consideration of All Binarized Convolutional Neural Network |
Sub Title (in English) | |
Keyword(1) | Neural Network |
Keyword(2) | Convolutional Neural Network |
Keyword(3) | Binary Convolutional Neural Network |
1st Author's Name | Masayuki Shimoda |
1st Author's Affiliation | Tokyo Institude of Technology(Tokyo Inst. of Tech.) |
2nd Author's Name | Tomoya Fujii |
2nd Author's Affiliation | Tokyo Institude of Technology(Tokyo Inst. of Tech.) |
3rd Author's Name | Haruyoshi Yonekawa |
3rd Author's Affiliation | Tokyo Institude of Technology(Tokyo Inst. of Tech.) |
4th Author's Name | Shimpei Sato |
4th Author's Affiliation | Tokyo Institude of Technology(Tokyo Inst. of Tech.) |
5th Author's Name | Hiroki Nakahara |
5th Author's Affiliation | Tokyo Institude of Technology(Tokyo Inst. of Tech.) |
Date | 2017-07-27 |
Paper # | CPSY2017-28 |
Volume (vol) | vol.117 |
Number (no) | CPSY-153 |
Page | pp.pp.131-136(CPSY), |
#Pages | 6 |
Date of Issue | 2017-07-19 (CPSY) |