Electrical Tunability in Band-to-Band-Tunneling based Neuron for Low Power Neuromorphic Computing

Source
9th IEEE Electron Devices Technology and Manufacturing Conference Shaping the Future with Innovations in Devices and Manufacturing Edtm 2025
Date Issued
2025-01-01
Author(s)
Patil, Shubham
Sakhuja, Jayatika
Biswas, Anmol
Hajare, Hemant
Kadam, Abhishek
Deshmukh, Shreyas
Singh, Ajay Kumar
Lashkare, Sandip 
Ganguly, Udayan
DOI
Abstract
In this work, we show the electrical control in the ultra-energy and area-efficient BTBT-based Si-neuron and the impact on network performance. We show the control of gate bias and current threshold on the spiking threshold and frequency. Finally, we show the impact of such design space on SNN performance and a 10-layer spiking Convolutional Neural Network (CNN). The result demonstrates that neurons' post-fabrication electrical tuning capability is essential for SNN performance improvement.
Unpaywall
URI