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

2:00 PM - 2:15 PM JST(5:00 AM - 5:15 AM UTC)

[8p-A23-1] [The 60th Young Scientist Presentation Award Speech] Dielectric Metasurface-based Hybrid Optical Neural Networks with Enhanced Generalizability via Representation Learning

〇Deyu Cao¹, Chun Ren¹, Chiyun Li¹, Andre Rohm², Ryoichi Horisaki², Takuo Tanemura¹ (1.Graduate School of Eng., UTokyo, 2.Graduate School of IST, UTokyo)

Keywords:

optical neural networks,deep learning

A pretraining framework using representation learning is introduced to enhance the generalizability of metasurface-based hybrid optical neural networks. High classification accuracy, outperforming digital baselines, is obtained numerically and experimentally for diverse datasets using a0.49-mm² metasurface.