Neuromorphic Computing for AI at the Edge

Cory Merkel, Ph.D.

Assistant Professor, Department of Computer Engineering, RIT

2/17/2023

11AM – 12PM CST

https://utsa.webex.com/utsa/j.php?MTID=m28e907e881c8d212904a7a90902e92cc

Abstract:  Artificial intelligence (AI) has become the linchpin in a growing number of products, services, and research programs which are aimed at automating and enhancing the human decision-making process.  There is no doubt that AI will play a central role in the future of healthcare, transportation, manufacturing, and defense, among others.  However, the rapidly growing size, weight, and power (SWaP) cost of AI algorithms limits their deployment on devices with practical computing and energy constraints (satellites, wearables, wireless sensors).  In this talk, I will discuss our lab’s research to bridge this gap and enable AI in the most SWaP-constrained environments.  This research takes a holistic approach, examining the entire AI stack, from devices and circuits to algorithms and applications.  At the lowest level, I will present our research on memristor-based circuits for implementing weighted communication pathways in artificial neural networks (ANNs).  Memristors reduce the power and latency associated with running ANNs on traditional computer architectures by directly emulating both the memory and computation of biological synapses.  In addition, memristor plasticity enables on-chip learning and allows ANNs to function in the presence of hardware defects and process variations.  Moving up the design hierarchy, I will discuss research on ANN topologies with partially random connectivity, which can lead to reduced hardware overhead and training cost while achieving state-of-the-art performance on classification tasks.  Finally, the talk will highlight some recent research related to the trustworthiness and potential security vulnerabilities of AI hardware.