[15a-M_123-1~13] FS.1 Focused Session "AI Electronics"

9:30 AM - 9:45 AM JST(12:30 AM - 12:45 AM UTC)

[15a-M_123-3] Power Consumption Evaluation of Weight Quantized SNN Circuits using Resistive Synaptic Devices

〇(M1)Ayane Matsuzaki¹, Rei Kusunose¹, Seiji Adachi¹, Kota Ando¹, Tetsuya Asai¹, Takao Marukame¹ (1.Hokkaido Univ.)

Keywords:

Spiking Neural Network,Quantization,Circuit

In this study, we evaluated the impact of weight quantization on power consumption in SNN circuits using Resistive RAM (ReRAM) and LIF neurons via simulations assuming a 22 nm process. The analysis revealed that reducing weight bit-width significantly suppresses the power increase associated with higher synapse counts and contributes to reducing ReRAM leakage power. Consequently, we demonstrated that extremely low-energy operation for the entire system, including energy per spike, is achievable.