[F-6] Charge-Based Memory Devices for AI Applications

[F-6-04] Implementation of Homeostasis Functionality in Hardware-Based Spiking Neural Networks Using an STDP Learning Rule

〇Jangsaeng Kim¹, Woo Young Choi¹, Byung-Gook Park¹, Jong-Ho Lee¹ (1. Seoul National University (Korea))

Presentation style: On-site (in-person)

https://doi.org/10.7567/SSDM.2022.F-6-04

We propose the two implementation methods of homeostasis functionality in hardware-based spiking neural networks (SNNs). The performance of the proposed implementation methods was evaluated through the MNIST dataset classification. Both implementation methods show high performance and a significantly higher recognition rate (~92%) than when homeostasis functionality was not used (~80%). The TFT-type flash memory cells are used as synaptic devices.