Enabling Spiking Neuromorphic Computation with On-Board Learning Through Algorithm and Hardware Co-design

Project Description

As AI becomes more prevalent in everyday applications, there is a need for new algorithms and dedicated chips that allow for low‐power, reliable and fast computing. In this project, we tackle this challenge by taking inspiration from spiking neural networks in nature. We take a distributed and biologically plausible learning rule that our team discovered recently, and computationally simulate its behaviour on networks involving new electrical components known as memristors. The novelty of our approach lies in the simplicity of our learning rule, its powerful ability to learn sequences and time series, and its robustness to hardware errors and noise in the training signals.

Research Technical Area

  • Cognitive modelling and systems
  • Machine learning
  • Neuromorphic computing

Benefits to the society

When the power consumption of AI chips become low enough to be embedded in mobile phones, sensors and household appliances, we will start to see an exponential explosion in the variety of ways that these devices serve us and communicate with us. For instance, our microwaves and refrigerators will be smart enough to understand our speech without sending the audio data to the cloud for processing.

Project’s Publications

Team’s Principal Investigator

Assistant Professor Shaowei Lin
Engineering Systems and Design Pillar
Singapore University of Technology and Design


Shaowei received his Ph.D. in Mathematics in 2011 from the University of California, Berkeley, where he analysed singularities in statistical models over large data sets through the lens of modern algebraic geometry. This work was continued at Stanford University in a one‐year DARPA postdoctoral collaboration with Andrew Ng to explore mathematical challenges in deep learning. In 2012, he returned to Singapore to start the Sense‐making Group in the Sense and Sense‐abilities programme in A*STAR. The group focused on exploiting deep learning techniques in sensor systems to enable intelligence at the edge of the network. In 2016, Shaowei joined SUTD where he crystallised his ideas for Distributed Artificial Intelligence. His research focuses on biologically plausible local learning rules for spiking neural networks based on path integrals in statistical physics, and on hierarchical geometric inference rules for scalable machine reasoning based on homotopy-type theory. In 2017, he spent six months as a visiting professor at Tomaso Poggio’s Center for Brains, Minds and Machines at MIT. Shaowei is currently an assistant professor in the Engineering Systems and Design pillar at SUTD.

Recent Notable Awards

  • 2015 MTI Borderless Silver Award MTI
  • 2014 Finalist at World Smart Cities Award
  • 2014 A*STAR TALENT Award

The Team

Co-Principal Investigator

Assoc. Prof. Zhao Rong, Singapore University of Technology and Design

Research Focus: Neuromorphic device and chip

Wang Chao, Senior Research Fellow, SUTD