[17p-WL1_301-1~13] Neuromorphic computing at the convergence of biology, electronics, and photonics

5:55 PM - 6:10 PM JST(8:55 AM - 9:10 AM UTC)Presentation by Applicant for JSAP Young Scientists Presentation Award

[17p-WL1_301-12] Brain-inspired computing based on photonic hyperdimensional vector generation

〇Takuya Iwata¹, Tomoaki Niiyama¹, Satoshi Sunada¹ (1.Kanazawa Univ.)

Keywords:

Optical Computing,Hyperdimensional Computing

In recent years, research on optical neural networks (ONNs) has been gaining momentum to accelerate AI processing. However, these networks face challenges such as the low noise tolerance inherent in analog computing and high power consumption due to frequent photoelectric conversion. In this study, we propose a photonic implementation of Hyperdimensional Computing (HDC), a brain-inspired computational model, to address these issues. The proposed optical HDC achieves high noise robustness by distributing information across high-dimensional vectors (HVs) and enables learning and inference using only simple vector operations while minimizing photoelectric conversion. In this paper, we report on the proof-of-concept for parallel optical HV generation and all-optical inference using a hybrid system of photonic integrated circuits and free-space optics.