Neuromorphic Computing Unleashing Intelligence with Single-Transistor SNNs

Abstract

Today, the importance of Artificial Intelligence (AI) is being appreciated as never before. A neural network can be considered as a brain that can be trained and has the ability to self-learn things. Hence, it allows the machines to be able to control their decisions. It is a well-known fact that future of technological world lies in AI. Therefore, all developed and most of the developing countries focus on research and innovation in AI. AI plays a key role in automation industries in terms of quality assurance, accuracy and has benefits primarily in terms of cost and time. To solve the problems of the society in a broader sense, neural networks must be introduced in more important fields like agriculture, education, healthcare and environmental sector. Since conventional systems require larger resources to operate, it becomes quiet challenging to cover these sectors under AI and neural networks. In this regard, the role of low cost, low power and area efficient hardware devices becomes inevitable. Inspired by human brain, neuromorphic computational architectures are based on neural networks. The third generation of neural networks is known as Spiking Neural Networks (SNNs) that work on discrete spikes. Making use of the mathematical models of biological neurons, several structures have been proposed that mimic the spiking behaviour of a biological neuron. The huge explosion in the field of technology has opened the way to concentrate on the practical aspects of these models. In the field of AI, stress is laid upon Speech and Image recognition and solving complex engineering problems. To solve the practical problems in a much broader sense, it becomes necessary to develop hardware based models of the biological neuron. With the advancement in VLSI technology, continuous efforts are being made to design hardware based Spiking Neural Networks that bridge the gap between Biological Neural Networks (BNNs) and ANNs. This talk will illustrate some extensions to this concept primarily focusing on SNNs using a single transistor as the fundamental building block, highlighting its transformative potential. These concepts will exhibit the need for synergy and careful co-design between devices and circuits to pave the way for the next-generation electronics.

Bio

FAISAL BASHIR received the master’s degrees in Electronic & IT from the University of Kashmir, Srinagar, India, and the Ph.D. degree in Applied Science/Electronics and Communication Engineering from Jami Millia Islamia, New Delhi, India. Besides this, he has qualified National Eligibility Test in electronic science for Assistant Professorship conducted by university grants commission. His research interests include Third generation neural networks, Biosensing and Nano Semiconductor device and modelling. He is currently working as an Assistant Professor with the Department of Computer Engineering and College of Computer Science and Information Technology, King Faisal University Hofuf, Al-Ahsa, Saudi Arabia. He has authored or co-authored more than 50 research publication and three book chapters in the field of Semiconductor devices, Neuromorphic Computing and VLSI design in reputed international journals and conferences. Dr. Faisal was the recipient of Best Paper Award in World Congress on Engineering held at Honk Kong 2014, Recipient of best presenter award at International Virtual Conference on Artificial Intelligence for Smart Community (IVC – AISC 2020) organised by Universiti Teknologi PETRONAS and Merit Scholarship at M.Phil. Level from the University of Kashmir.