Proceedings of MATSUS Spring 2025 Conference (MATSUSSpring25)
Publication date: 16th December 2024
The rapid advancement of Internet of Things (IoT) and artificial intelligence (AI) technologies has driven extensive research into artificial neural networks (ANNs) that mimic the synaptic characteristics of the human brain. These systems offer advantages such as low power consumption, high integration, and parallel computing. Recently, hardware-based ANNs utilizing arrays of electronic devices—where synaptic devices serve as synaptic weights—have demonstrated high accuracy in real-time detection and information processing. However, they face critical challenges, including slow processing speeds and high energy consumption due to electrical interference and low uniformity among array cells during the learning process.
To address these limitations, we propose an optical signal-based system capable of analog conductance control with high uniformity and low energy consumption. By leveraging a polymer film with tunable conductance—adjustable through light intensity and exposure time—we demonstrate a neural network that enables continuous conductance modulation for improved image recognition accuracy, eliminating electrical interference during the writing process. This innovative low-power and reliable image recognition system presents a promising platform for advanced image processing applications.
This research was supported by the National Research Foundation of Korea funded by the Ministry of Science, ICT & Future Planning (grant no.NRF-2022M3C1A3081211)