[18p-P06-1~11] FS Focused Session "AI Electronics" (Poster)

[18p-P06-3] Comparison on Performance for Optical Convolutional Neural Networks and Photonic Reservoir Computing Using EO polymer/Si Hybrid Modulator

〇Akito Shinya¹, Guo-Wei Lu², Koji Kida³, Hiromu Sato², Shiyoshi Yokoyama², Junichi Fujikata¹ (1.Tokushima Univ., 2.Kyushu Univ., 3.Kagawa Univ.)

Keywords:

Photonic Reservoir Computing,Optical Convolutional Neural Network,Hybrid Optical Modulator

Photonic reservoir computing (Photonic RC) is one of the recurrent neural networks (RNN), which is characterized by the fact that it learns in the output layer by fixing the combined weights of the input data and reservoir layer and the feedback weights in the reservoir layer. Therefore, it is expected to be suitable for physical implementation due to its relatively simple computation system configuration, and to be able to perform optical operations at high speed and low power consumption. In this study, an optical convolutional neural network (OCNN) with an EO polymer/Si hybrid optical modulator and an photonic RC with integrated nonlinear waveguide were examined and compared for image recognition accuracy and computation time.