| PhD Viva

Name of the Speaker: Mr. Shubham Pande (EE18D704)
Guide: Dr. Bhaswar Chakrabarti
Co-Guide: Dr. Anjan Chakravorty
Online meeting link: https://meet.google.com/met-oiqe-oou
Date/Time: 20th March 2025 (Thursday), 10:00 AM
Title: ReRAM Memory Technology for Neuromorphic Computing: Design, Modeling and Applications

Abstract :

This thesis explores the use of Resistive Random-Access Memory (ReRAM) in neuromorphic computing, focusing on thermal effects, switching energy, and spiking neural networks (SNNs) training. It first develops a physics-based compact model for thermal resistance and crosstalk to ensure accurate SPICE simulations. Additionally, it examines electrothermal effects and shows that switching energy can be reduced by trapping heat using a thermal barrier layer at the electrode-switching oxide interface. The final phase applies ReRAM to SNNs, which, despite their biological inspiration, struggle with accuracy and energy efficiency. To overcome these challenges, this thesis introduces “neoHebbian synapses,” which leverage temperature as an additional state variable for efficient online learning. These synapses accelerate SNN training and improve performance. Overall, this work advances the modeling, design, and application of ReRAM, contributing to the development of energy-efficient, high-performance neuromorphic computing systems.