Abstract

In today’s computing systems based on the conventional von Neumann architecture, there are distinct memory and processing units. Performing computations results in a significant amount of data being moved back and forth between the physically separated memory and processing units. This costs time and energy and constitutes an inherent performance bottleneck. It is becoming increasingly clear that for AI application areas, we need to transition to computing architectures in which memory and logic coexist in some form. Brain-inspired neuromorphic computing and the fascinating new area of in-memory computing or computational memory are two key non-von Neumann approaches being researched. A critical requirement in these novel computing paradigms is a very-high-density, low-power, variable-state, programmable and non-volatile nanoscale memory device. There are many examples of such nanoscale memory devices in which the information is stored either as charge or as resistance. However, one particular example is phase-change-memory (PCM) devices, which are very well suited to address this need, owing to their multi-level storage capability and potential scalability.

In in-memory computing, the physics of the nanoscale memory devices, as well as the organization of such devices in crossbar arrays, are exploited to perform certain computational tasks within the memory unit. We will present how computational memories accelerate AI applications and will show small- and large-scale experimental demonstrations that perform high-level computational primitives, such as ultra-low-power inference engines, optimization solvers including compressed sensing and sparse coding, linear solvers and temporal correlation detection. Moreover, we will discuss the efficacy of this approach to efficiently address not only inferencing but also training of deep neural networks. In neuromorphic computing, we explore power- and area-efficient hardware architectures for learning systems that are inspired by the structure and the functioning principles of the brain. We will present how the biologically-feasible spiking neurons and synapses can be implemented in an all-PCM architecture and will show experimental demonstrations of unsupervised and supervised learning. The results of in-memory and neuromorphic computing show that this co-existence of computation and storage at the nanometer scale could be the enabler for new, ultra-dense, low-power, and massively parallel computing systems. Thus, by augmenting conventional computing systems, in-memory and neuromorphic computing could help achieve orders of magnitude improvement in performance and efficiency.

Short Bio Dr. Eleftheriou

Evangelos Eleftheriou, received his Ph.D. degree in Electrical Engineering from Carleton University, Ottawa, Canada, in 1985. In 1986, he joined the IBM Research – Zurich laboratory in Rüschlikon, Switzerland, as a Research Staff Member. Since 1998, he has held various management positions and is currently responsible for the neuromorphic computing activities of IBM Research – Zurich. His research interests include signal processing, coding, non-volatile memory technologies and emerging computing paradigms such as neuromorphic and in-memory computing for AI applications. He has authored or coauthored over 200 publications, and holds over 160 patents (granted and pending applications). In 2002, he became a Fellow of the IEEE. He was co-recipient of the 2003 IEEE Communications Society Leonard G. Abraham Prize Paper Award. He was also co-recipient of the 2005 Technology Award of the Eduard Rhein Foundation. In 2005, he was appointed an IBM Fellow. The same year he was also inducted into the IBM Academy of Technology. In 2009, he was co-recipient of the IEEE CSS Control Systems Technology Award and of the IEEE Transactions on Control Systems Technology Outstanding Paper Award. In 2016, he received an honoris causa professorship from the University of Patras, Greece. In 2018, he was inducted into the US National Academy of Engineering as Foreign Member.

Short Bio Dr. Pantazi

Angeliki Pantazi is a Research Staff Member at the IBM Research – Zurich in Switzerland. She received her Diploma and Ph.D. degrees in Electrical Engineering and Computer Technology from the University of Patras, Greece, in 1996 and 2005, respectively. In 2002, she joined IBM Research – Zurich as a Ph.D. student and became a Research Staff Member in 2006. She was named IBM Master Inventor in 2014 and became a senior member of the IEEE in 2015. She was a co-recipient of the 2009 IEEE Control Systems Technology Award, the 2009 IEEE Transactions on Control Systems Technology Outstanding Paper Award and the 2014 IFAC Industrial Achievement Award. In 2017, she received an IBM Corporate Award and the IEEE Control Systems Society Transition to Practice Award. Her research interests include multiple control-related aspects of data storage systems, where she particularly contributed in the research and development of advanced servo control technologies for magnetic tape drive systems. Recently, her research is also focusing on neuromorphic technologies combined with emerging memory concepts such as phase-change memory. She has published more than 90 refereed articles and holds over 40 granted and pending patents.