[9p-PB6-1~21] 3.2 Information photonics and image engineering

[9p-PB6-21] All-Optical Nonlinear Activation Function Based on TPPS

〇Jin Nakamura¹, Kazuaki Nakata¹, Takayoshi Kobayashi², Eiji Tokunaga¹ (1.Tokyo Univ. of Sci., 2.National Yang Ming Chiao Tung Univ.)

Keywords:

activation function,J-aggregates,optical neural network

We performed an in silico evaluation of a system utilizing the saturable absorption of TPPS as an activation function for ONNs, demonstrating that it achieves an accuracy comparable to that of ReLU and sigmoid functions. Furthermore, we showed that measurement conditions—such as sample concentration, pH value, optical path length, and laser wavelength—can be determined by optimizing hyperparameters like the linear and nonlinear absorption coefficients.