End-to-End Optimization for a Compact Optical Neural Network Based on Nanostructured 2 × 2 Optical Processors
Recent research in silicon photonic chips has made huge progress in optical computing owing to their high speed, small footprint, and low energy consumption. Here, we employ nanostructured 2 × 2 optical processors in an optical neural network for implementing a binary classification task...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2023-01-01
|
| Series: | IEEE Photonics Journal |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10234101/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850093787960836096 |
|---|---|
| author | Caiyue Zhao Jiguang Wang Simei Mao Xuanyi Liu Wai Kin Victor Chan H. Y. Fu |
| author_facet | Caiyue Zhao Jiguang Wang Simei Mao Xuanyi Liu Wai Kin Victor Chan H. Y. Fu |
| author_sort | Caiyue Zhao |
| collection | DOAJ |
| description | Recent research in silicon photonic chips has made huge progress in optical computing owing to their high speed, small footprint, and low energy consumption. Here, we employ nanostructured 2 × 2 optical processors in an optical neural network for implementing a binary classification task efficiently. The proposed optical neural network is composed of five linear layers including ten optical processors in each layer, and nonlinear activation functions. 2 × 2 optical processors are designed based on digitized meta-structures which have an extremely compact footprint of 1.6 × 4 μm<sup>2</sup>. A brand-new end-to-end design strategy based on Deep Q-Network is proposed to optimize the optical neural network for classifying a generated ring data set with better generalization, robustness, and operability. A high-efficient transfer matrix multiplication method is applied to simplify the calculation process in traditional optical software. Our numerical results illustrate that the maximum and mean accuracy on the testing data set can reach 90.5% and 87.8%, respectively. The demonstrated optical processors with a significantly compact area, and the efficient optimization method exhibit high potential for large-scale integration of whole-passive optical neural network on a photonic chip. |
| format | Article |
| id | doaj-art-21eebf79ae6a4d47a0c7521018a96f8b |
| institution | DOAJ |
| issn | 1943-0655 |
| language | English |
| publishDate | 2023-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Photonics Journal |
| spelling | doaj-art-21eebf79ae6a4d47a0c7521018a96f8b2025-08-20T02:41:49ZengIEEEIEEE Photonics Journal1943-06552023-01-011551810.1109/JPHOT.2023.330983510234101End-to-End Optimization for a Compact Optical Neural Network Based on Nanostructured 2 × 2 Optical ProcessorsCaiyue Zhao0https://orcid.org/0000-0003-3832-9824Jiguang Wang1https://orcid.org/0009-0005-3034-1871Simei Mao2https://orcid.org/0000-0002-0486-9171Xuanyi Liu3https://orcid.org/0000-0001-7542-3882Wai Kin Victor Chan4https://orcid.org/0000-0002-7202-1922H. Y. Fu5https://orcid.org/0000-0002-4276-0011Tsinghua-Berkeley Shenzhen Institute and Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, ChinaTsinghua-Berkeley Shenzhen Institute and Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, ChinaTsinghua-Berkeley Shenzhen Institute and Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, ChinaTsinghua-Berkeley Shenzhen Institute and Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, ChinaTsinghua-Berkeley Shenzhen Institute and Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, ChinaTsinghua-Berkeley Shenzhen Institute and Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, ChinaRecent research in silicon photonic chips has made huge progress in optical computing owing to their high speed, small footprint, and low energy consumption. Here, we employ nanostructured 2 × 2 optical processors in an optical neural network for implementing a binary classification task efficiently. The proposed optical neural network is composed of five linear layers including ten optical processors in each layer, and nonlinear activation functions. 2 × 2 optical processors are designed based on digitized meta-structures which have an extremely compact footprint of 1.6 × 4 μm<sup>2</sup>. A brand-new end-to-end design strategy based on Deep Q-Network is proposed to optimize the optical neural network for classifying a generated ring data set with better generalization, robustness, and operability. A high-efficient transfer matrix multiplication method is applied to simplify the calculation process in traditional optical software. Our numerical results illustrate that the maximum and mean accuracy on the testing data set can reach 90.5% and 87.8%, respectively. The demonstrated optical processors with a significantly compact area, and the efficient optimization method exhibit high potential for large-scale integration of whole-passive optical neural network on a photonic chip.https://ieeexplore.ieee.org/document/10234101/Digitized meta-structureoptical computingoptical neural networksreinforcement learning |
| spellingShingle | Caiyue Zhao Jiguang Wang Simei Mao Xuanyi Liu Wai Kin Victor Chan H. Y. Fu End-to-End Optimization for a Compact Optical Neural Network Based on Nanostructured 2 × 2 Optical Processors IEEE Photonics Journal Digitized meta-structure optical computing optical neural networks reinforcement learning |
| title | End-to-End Optimization for a Compact Optical Neural Network Based on Nanostructured 2 × 2 Optical Processors |
| title_full | End-to-End Optimization for a Compact Optical Neural Network Based on Nanostructured 2 × 2 Optical Processors |
| title_fullStr | End-to-End Optimization for a Compact Optical Neural Network Based on Nanostructured 2 × 2 Optical Processors |
| title_full_unstemmed | End-to-End Optimization for a Compact Optical Neural Network Based on Nanostructured 2 × 2 Optical Processors |
| title_short | End-to-End Optimization for a Compact Optical Neural Network Based on Nanostructured 2 × 2 Optical Processors |
| title_sort | end to end optimization for a compact optical neural network based on nanostructured 2 x00d7 2 optical processors |
| topic | Digitized meta-structure optical computing optical neural networks reinforcement learning |
| url | https://ieeexplore.ieee.org/document/10234101/ |
| work_keys_str_mv | AT caiyuezhao endtoendoptimizationforacompactopticalneuralnetworkbasedonnanostructured2x00d72opticalprocessors AT jiguangwang endtoendoptimizationforacompactopticalneuralnetworkbasedonnanostructured2x00d72opticalprocessors AT simeimao endtoendoptimizationforacompactopticalneuralnetworkbasedonnanostructured2x00d72opticalprocessors AT xuanyiliu endtoendoptimizationforacompactopticalneuralnetworkbasedonnanostructured2x00d72opticalprocessors AT waikinvictorchan endtoendoptimizationforacompactopticalneuralnetworkbasedonnanostructured2x00d72opticalprocessors AT hyfu endtoendoptimizationforacompactopticalneuralnetworkbasedonnanostructured2x00d72opticalprocessors |