XU Xiaofeng,ZHULianqing,ZHUANGWei,ZHANGDongliang,LULidan,YUANPei.Optimization of optical convolution kernel of optoelectronic hybrid convolution neural network[J].Optoelectronics Letters,2022,(3):181-186 |
Optimization of optical convolution kernel of optoelectronic hybrid convolution neural network |
Author Name | Affiliation | XU Xiaofeng | School of Electro-Optical Engineering, Changchun University of Science and Technology, Jilin 130022, China | ZHULianqing | School of Electro-Optical Engineering, Changchun University of Science and Technology, Jilin 130022, China Key Laboratory of the Ministry of Education for Optoelectronic Measurement Technology and Instrument, Beijing Information Science and Technology University, Beijing 100192, China | ZHUANGWei | Key Laboratory of the Ministry of Education for Optoelectronic Measurement Technology and Instrument, Beijing Information Science and Technology University, Beijing 100192, China | ZHANGDongliang | Key Laboratory of the Ministry of Education for Optoelectronic Measurement Technology and Instrument, Beijing Information Science and Technology University, Beijing 100192, China | LULidan | Beijing Laboratory of Optical Fiber Sensing and System, Beijing Information Science and Technology University, Beijing 100016, China | YUANPei | Beijing Laboratory of Optical Fiber Sensing and System, Beijing Information Science and Technology University, Beijing 100016, China |
|
Abstract: |
To enhance the optical computation's utilization efficiency, we develop an optimization method for optical convolution kernel in the optoelectronic hybrid convolution neural network (OHCNN). To comply with the actual calculation process, the convolution kernel is expanded from single-channel to two-channel, containing positive and negative weights. The Fashion-MNIST dataset is used to test the network architecture's accuracy, and the accuracy is improved by 7.5% with the optimized optical convolution kernel. The energy efficiency ratio (EER) of two-channel network is 46.7% higher than that of the single-channel network, and it is 2.53 times of that of traditional electronic products. |
Hits: 355 |
Download times: 0 |
|
View Full Text Download reader |
|
|
|