Slimmed Optical Neural Networks with Multiplexed Neuron Sets and a Corresponding Backpropagation Training Algorithm
Optical neural networks (ONNs) have recently attracted extensive interest as potential alternatives to electronic artificial neural networks, owing to their intrinsic capabilities in parallel signal processing with reduced power consumption and low latency. Preliminary confirmation of parallelism in...
Main Authors: | , , , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
American Association for the Advancement of Science (AAAS)
2024-01-01
|
Series: | Intelligent Computing |
Online Access: | https://spj.science.org/doi/10.34133/icomputing.0070 |
_version_ | 1797301241483624448 |
---|---|
author | Yi-Feng Liu Rui-Yao Ren Dai-Bao Hou Hai-Zhong Weng Bo-Wen Wang Ke-Jie Huang Xing Lin Feng Liu Chen-Hui Li Chao-Yuan Jin |
author_facet | Yi-Feng Liu Rui-Yao Ren Dai-Bao Hou Hai-Zhong Weng Bo-Wen Wang Ke-Jie Huang Xing Lin Feng Liu Chen-Hui Li Chao-Yuan Jin |
author_sort | Yi-Feng Liu |
collection | DOAJ |
description | Optical neural networks (ONNs) have recently attracted extensive interest as potential alternatives to electronic artificial neural networks, owing to their intrinsic capabilities in parallel signal processing with reduced power consumption and low latency. Preliminary confirmation of parallelism in optical computing has been widely performed by applying wavelength division multiplexing (WDM) to the linear transformation of neural networks. However, interchannel crosstalk has obstructed WDM technologies from being deployed in nonlinear activation on ONNs. Here, we propose a universal WDM structure called multiplexed neuron sets (MNS), which applies WDM technologies to optical neurons and enables ONNs to be further compressed. A corresponding backpropagation (BP) training algorithm was proposed to alleviate or even annul the influence of interchannel crosstalk in MNS-based WDM-ONNs. For simplicity, semiconductor optical amplifiers are employed as an example of MNS to construct a WDM-ONN trained using the new algorithm. The results show that the combination of MNS and the corresponding BP training algorithm clearly downsizes the system and improves the energy efficiency by a factor of 10 while providing similar performance to traditional ONNs. |
first_indexed | 2024-03-07T23:19:40Z |
format | Article |
id | doaj.art-23087f20f6ac422d83837bae838c2172 |
institution | Directory Open Access Journal |
issn | 2771-5892 |
language | English |
last_indexed | 2024-03-07T23:19:40Z |
publishDate | 2024-01-01 |
publisher | American Association for the Advancement of Science (AAAS) |
record_format | Article |
series | Intelligent Computing |
spelling | doaj.art-23087f20f6ac422d83837bae838c21722024-02-21T07:40:07ZengAmerican Association for the Advancement of Science (AAAS)Intelligent Computing2771-58922024-01-01310.34133/icomputing.0070Slimmed Optical Neural Networks with Multiplexed Neuron Sets and a Corresponding Backpropagation Training AlgorithmYi-Feng Liu0Rui-Yao Ren1Dai-Bao Hou2Hai-Zhong Weng3Bo-Wen Wang4Ke-Jie Huang5Xing Lin6Feng Liu7Chen-Hui Li8Chao-Yuan Jin9College of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027, China.College of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027, China.College of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027, China.School of Physics, CRANN and AMBER, Trinity College Dublin, Dublin 2, Ireland.Synopsys, Inc., 7521PL Enschede, the Netherlands.College of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027, China.College of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027, China.College of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027, China.College of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027, China.College of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027, China.Optical neural networks (ONNs) have recently attracted extensive interest as potential alternatives to electronic artificial neural networks, owing to their intrinsic capabilities in parallel signal processing with reduced power consumption and low latency. Preliminary confirmation of parallelism in optical computing has been widely performed by applying wavelength division multiplexing (WDM) to the linear transformation of neural networks. However, interchannel crosstalk has obstructed WDM technologies from being deployed in nonlinear activation on ONNs. Here, we propose a universal WDM structure called multiplexed neuron sets (MNS), which applies WDM technologies to optical neurons and enables ONNs to be further compressed. A corresponding backpropagation (BP) training algorithm was proposed to alleviate or even annul the influence of interchannel crosstalk in MNS-based WDM-ONNs. For simplicity, semiconductor optical amplifiers are employed as an example of MNS to construct a WDM-ONN trained using the new algorithm. The results show that the combination of MNS and the corresponding BP training algorithm clearly downsizes the system and improves the energy efficiency by a factor of 10 while providing similar performance to traditional ONNs.https://spj.science.org/doi/10.34133/icomputing.0070 |
spellingShingle | Yi-Feng Liu Rui-Yao Ren Dai-Bao Hou Hai-Zhong Weng Bo-Wen Wang Ke-Jie Huang Xing Lin Feng Liu Chen-Hui Li Chao-Yuan Jin Slimmed Optical Neural Networks with Multiplexed Neuron Sets and a Corresponding Backpropagation Training Algorithm Intelligent Computing |
title | Slimmed Optical Neural Networks with Multiplexed Neuron Sets and a Corresponding Backpropagation Training Algorithm |
title_full | Slimmed Optical Neural Networks with Multiplexed Neuron Sets and a Corresponding Backpropagation Training Algorithm |
title_fullStr | Slimmed Optical Neural Networks with Multiplexed Neuron Sets and a Corresponding Backpropagation Training Algorithm |
title_full_unstemmed | Slimmed Optical Neural Networks with Multiplexed Neuron Sets and a Corresponding Backpropagation Training Algorithm |
title_short | Slimmed Optical Neural Networks with Multiplexed Neuron Sets and a Corresponding Backpropagation Training Algorithm |
title_sort | slimmed optical neural networks with multiplexed neuron sets and a corresponding backpropagation training algorithm |
url | https://spj.science.org/doi/10.34133/icomputing.0070 |
work_keys_str_mv | AT yifengliu slimmedopticalneuralnetworkswithmultiplexedneuronsetsandacorrespondingbackpropagationtrainingalgorithm AT ruiyaoren slimmedopticalneuralnetworkswithmultiplexedneuronsetsandacorrespondingbackpropagationtrainingalgorithm AT daibaohou slimmedopticalneuralnetworkswithmultiplexedneuronsetsandacorrespondingbackpropagationtrainingalgorithm AT haizhongweng slimmedopticalneuralnetworkswithmultiplexedneuronsetsandacorrespondingbackpropagationtrainingalgorithm AT bowenwang slimmedopticalneuralnetworkswithmultiplexedneuronsetsandacorrespondingbackpropagationtrainingalgorithm AT kejiehuang slimmedopticalneuralnetworkswithmultiplexedneuronsetsandacorrespondingbackpropagationtrainingalgorithm AT xinglin slimmedopticalneuralnetworkswithmultiplexedneuronsetsandacorrespondingbackpropagationtrainingalgorithm AT fengliu slimmedopticalneuralnetworkswithmultiplexedneuronsetsandacorrespondingbackpropagationtrainingalgorithm AT chenhuili slimmedopticalneuralnetworkswithmultiplexedneuronsetsandacorrespondingbackpropagationtrainingalgorithm AT chaoyuanjin slimmedopticalneuralnetworkswithmultiplexedneuronsetsandacorrespondingbackpropagationtrainingalgorithm |