The Japan Society of Applied Physics

2023 International Conference on Solid State Devices and Materials

2023 International Conference on Solid State Devices and Materials

Sep 5, 2023 - Oct 31, 2023
Nagoya Congress Center

3:45 PM - 4:00 PM

[K-8-02] Hardware Implementation of Temporal Encoded Spiking Neural Networks and Its Application to Rapid Gas Detection.

○Jiseong Im1,2, Donghee Kim1,2, Woo Young Choi1,2, Jong-Ho Lee1,2 (1. Seoul National University (Korea), 2. Inter-Univ. Semiconductor Res. Center (ISRC) (Korea))

https://doi.org/10.7567/SSDM.2023.K-8-02

A hardware implementation of the temporal encoded spiking neural network is proposed. Fabricated TFTs with 20 precise conductance levels and designed circuits for temporal backpropagation are used. FET-type gas sensors with an In2O3 sensing layer are fabricated, and their measured data are used for training the proposed network. As a result of the high-level simulation under ideal device condition, the network predicts the type and concentration of gas fast (~4.3 s) with a small error (< 2%).