[8p-A23-1~12] CS.2 Code-sharing session of 3.2&4.4

2:30 PM - 2:45 PM JST(5:30 AM - 5:45 AM UTC)Presentation by Applicant for JSAP Young Scientists Presentation Award

[8p-A23-3] Hybrid Optical Neural Networks with Simultaneous Perturbation Stochastic Approximation

〇(M1)Kengo Takahashi¹, Hironori Ito¹, Satoshi Honma¹ (1.U. Yamanashi Grad.)

Keywords:

Optical neural network,Simultaneous Perturbation Stochastic Approximation

Hybrid optical neural networks (ONNs) integrating optical computing and electronic processing have attracted significant attention. However, training based on backpropagation requires prior measurement and modeling of the optical system's transfer characteristics. In this report, we propose a hybrid ONN trained via simultaneous perturbation stochastic approximation (SPSA), which estimates gradients directly from the physical output. We verify the feasibility of a system combining a multi-level nematic liquid crystal spatial light modulator (NLC-SLM) and a high-speed binary ferroelectric liquid crystal SLM (FLC-SLM).