A 41.3pJ/26.7pJ Per Neuron Weight RBM Processor for on-Chip Learning/Inference Applications
顯示簡易詮釋資料 統計資料| 欄位 | 內容 | 語言 |
|---|---|---|
dc.citation.epage |
268 | en_US |
dc.citation.spage |
265 | en_US |
dc.contributor.author |
Tsai, Chang-Hung | en_US |
dc.contributor.author |
YU, Wan-Jun | en_US |
dc.contributor.author |
Wong, Wing Hung | en_US |
dc.contributor.author |
Lee, Chen-Yi | en_US |
dc.contributor.department |
電子工程學系及電子研究所 | zh_TW |
dc.contributor.department |
Department of Electronics Engineering and Institute of Electronics | en_US |
dc.date.accessioned |
2018-08-21T05:56:43Z | |
dc.date.available |
2018-08-21T05:56:43Z | |
dc.date.issued |
2016-01-01 | en_US |
dc.description.abstract |
A restricted Boltzmann machine (RBM) processor (RBM-P) supporting on-chip learning and inference is proposed for machine learning applications in this paper. Featuring neural network (NN) model reduction for external memory bandwidth saving, low power neuron binarizer (LPNB) with dynamic clock gating and area-efficient NN-like activation function calculators, user-defined connection map (UDCM) for both computation time and bandwidth saving, and early stopping (ES) mechanism in learning process, the proposed system integrates 32 RBM cores with maximal 4k neurons per layer and 128 candidates per sample for machine learning applications. Implemented in 65nm CMOS technology', the proposed RBM-P chip costs 2.2M gates and 128kB SRAM with 8.8mm(2) area. Operated at 1.2V and 210MHz, this chip respectively achieves 114.3x and 3.9x faster processing time than CPU and GPGPU. And the proposed RBM-P chip consumes 41.3pJ and 26.7pJ per neuron weight (NW) for learning and inference, respectively. | en_US |
dc.identifier.journal |
2016 IEEE ASIAN SOLID-STATE CIRCUITS CONFERENCE (A-SSCC) | en_US |
dc.identifier.uri |
https://ir.lib.nycu.edu.tw/handle/11536/146556 | |
dc.identifier.wosnumber |
WOS:000401471500067 | en_US |
dc.language.iso |
en_US | en_US |
dc.subject |
restricted Boltzmann machine (RBM) | en_US |
dc.subject |
machine learning | en_US |
dc.subject |
non-linear functions | en_US |
dc.subject |
low power design | en_US |
dc.title |
A 41.3pJ/26.7pJ Per Neuron Weight RBM Processor for on-Chip Learning/Inference Applications | en_US |
dc.type |
Proceedings Paper | en_US |
檔案
本條目沒有附加檔案。
所屬典藏: 會議論文
引用此條目
Tsai, C.-H., YU, W.-J., Wong, W. H., & Lee, C.-Y. (2016). A 41.3pJ/26.7pJ Per Neuron Weight RBM Processor for on-Chip Learning/Inference Applications. 2016 IEEE ASIAN SOLID-STATE CIRCUITS CONFERENCE (A-SSCC), 265–268. https://ir.lib.nycu.edu.tw/handle/11536/146556
引用文字依 APA 第 7 版格式自動產生,使用前請再核對。
分享
永久網址:https://ir.lib.nycu.edu.tw/handle/11536/146556