東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Architecture Design of Opto-Electron...

Zhang, Jiaxu,

FindBook

Google Book

Amazon

博客來

Architecture Design of Opto-Electronic Neural Computing System /

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Architecture Design of Opto-Electronic Neural Computing System // Jiaxu Zhang.
作者:	Zhang, Jiaxu,
面頁冊數:	1 electronic resource (127 pages)
附註:	Source: Dissertations Abstracts International, Volume: 86-05, Section: B.
Contained By:	Dissertations Abstracts International86-05B.
標題:	Design. -
電子資源:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31718169
ISBN:	9798346542148

Architecture Design of Opto-Electronic Neural Computing System /
Zhang, Jiaxu,

Architecture Design of Opto-Electronic Neural Computing System /Jiaxu Zhang. - 1 electronic resource (127 pages)

Source: Dissertations Abstracts International, Volume: 86-05, Section: B.

Neural Networks (NNs) have been applied to various scenarios in people life and played a significant role. However, for conventional digital computing systems, the limitations of the Von-Neumann architecture and the ending of Moore's Law pose bottlenecks of performance and power efficiency in NN training and inference. To address these challenges, Optical Neural Networks (ONNs) based on the silicon photonic platform have emerged. These architectures demonstrate high-bandwidth, low-latency, and power-efficient data processing in optical cores, while peripheral electronic subsystems face issues including speed bottlenecks, high costs, and significant power consumption.We surveyed and analyzed various optical computing architectures, and by comparing them with the electronic digital NN accelerators, the design ideas of our optoelectronic hybrid neural computing system are determined, including pipelined optical cores, highfrequency stream operations and optimizations of data transmission across data format domains. We designed and optimized UONN architecture, which starts from cost-efficient photonic cores, greatly reducing the data conversion pressure on peripheral systems by the asynchronous loop structure in the optical domain. We also designed the PHOENIX architecture that utilizes dedicated core structures for specific linear operations with pipeline optimized, and builds an "optical-analog-digital" three-level memory hierarchy. In addition, to evaluate and compare our works, we also designed a simulation platform for ONN that is oriented to the general ONN architecture. It can build a functional model of ONN system based on external configurations and perform behavioral simulation of task execution, to obtain evaluation results such as performance and energy. The systems we designed achieves several orders of magnitude lead in performance, energy consumption and computing density compared to other electronic/optical NN accelerators.

English

ISBN: 9798346542148Subjects--Topical Terms:

518875
Design.

Architecture Design of Opto-Electronic Neural Computing System /
LDR:03246nmm a22003853i 4500 001 2434973
005 20250930135653.5
006 m o d
007 cr|nu||||||||
008 261011s2024 miu||||||m |||||||eng d
020 $a 9798346542148
035 $a (MiAaPQD)AAI31718169
035 $a AAI31718169
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Zhang, Jiaxu, $e author. $3 3825144
245 1 0 $a Architecture Design of Opto-Electronic Neural Computing System / $c Jiaxu Zhang.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2024
300 $a 1 electronic resource (127 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 86-05, Section: B.
500 $a Advisors: Xu, Jiang; Zhang, Wei.
502 $b Ph.D. $c Hong Kong University of Science and Technology (Hong Kong) $d 2024.
520 $a Neural Networks (NNs) have been applied to various scenarios in people life and played a significant role. However, for conventional digital computing systems, the limitations of the Von-Neumann architecture and the ending of Moore's Law pose bottlenecks of performance and power efficiency in NN training and inference. To address these challenges, Optical Neural Networks (ONNs) based on the silicon photonic platform have emerged. These architectures demonstrate high-bandwidth, low-latency, and power-efficient data processing in optical cores, while peripheral electronic subsystems face issues including speed bottlenecks, high costs, and significant power consumption.We surveyed and analyzed various optical computing architectures, and by comparing them with the electronic digital NN accelerators, the design ideas of our optoelectronic hybrid neural computing system are determined, including pipelined optical cores, highfrequency stream operations and optimizations of data transmission across data format domains. We designed and optimized UONN architecture, which starts from cost-efficient photonic cores, greatly reducing the data conversion pressure on peripheral systems by the asynchronous loop structure in the optical domain. We also designed the PHOENIX architecture that utilizes dedicated core structures for specific linear operations with pipeline optimized, and builds an "optical-analog-digital" three-level memory hierarchy. In addition, to evaluate and compare our works, we also designed a simulation platform for ONN that is oriented to the general ONN architecture. It can build a functional model of ONN system based on external configurations and perform behavioral simulation of task execution, to obtain evaluation results such as performance and energy. The systems we designed achieves several orders of magnitude lead in performance, energy consumption and computing density compared to other electronic/optical NN accelerators.
546 $a English
590 $a School code: 1223
650 4 $a Design. $3 518875
650 4 $a Energy efficiency. $3 3555643
650 4 $a Decision making. $3 517204
650 4 $a Neural networks. $3 677449
650 4 $a Sustainability. $3 1029978
690 $a 0389
690 $a 0800
690 $a 0640
710 2 $a Hong Kong University of Science and Technology (Hong Kong). $e degree granting institution. $3 3814653
720 1 $a Xu, Jiang $e degree supervisor.
720 1 $a Zhang, Wei $e degree supervisor.
773 0 $t Dissertations Abstracts International $g 86-05B.
790 $a 1223
791 $a Ph.D.
792 $a 2024
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31718169