Spiking Equilibrium Propagation for Real-time Learning

Implemented the Equi-Prop system using 45nm CMOS technology

Schematic for the Equi-Prop system

Guide: Prof. Udayan Ganguly

This project was taken up as part of the course EE746: Neuromorphic Engineering by Prof. Udayan Ganguly. The problem statement was to implement Spiking Equilibrium Propagation using the 45nm CMOS technology. The original paper had shown the CMOS implementation using a 65nm technology. Using our novel implementation of the Equi-Prop system in 45nm technology, we were able to simulate the synapse between two neurons. The circuit consumes a total power of 82.65uW and the synapse circuit only uses about 8.79uW of this.

     Report      Slides      GitHub

References

2022

  1. CMOS Implementation of Spiking Equilibrium Propagation for Real-Time Learning
    Brady Taylor, Nicky Ramos, Eric Yeats, and 1 more author
    In 2022 IEEE 4th International Conference on Artificial Intelligence Circuits and Systems (AICAS), May 2022

2017

  1. Equilibrium Propagation: Bridging the Gap between Energy-Based Models and Backpropagation
    Benjamin Scellier, and Yoshua Bengio
    Frontiers in Computational Neuroscience, May 2017