JUMP e-Workshop / Dynamical Systems for Graph Analytics and Optimization Problems


Location: webex

Speakers: Sourav Dutta / Suman Datta (University of Notre Dame)

Event Brief: In this talk, we will present a continuous-time dynamical system (CTDS) using coupled networks of stochastic phase-transition nano-oscillators (PTNOs) that can solve computationally hard graph analytics and combinatorial optimization problems with massive speed-up and energy-efficiency.

Abstract: The unprecedented growth in the volume of unstructured and dynamically changing data has led to a need for specialized hardware that can accelerate graph analytics. Representative applications include modeling supply-chains, spread of infection diseases, allocating resources under constraints and optimal MIMO detection for 5G networks. Many of such computationally hard problems belong to the non-deterministic polynomial time (NP)-hard complexity class, meaning they need an exponentially increasing amount of resources (time, energy, hardware) to find the solution as the problem size increases. Interestingly, a large class of such challenging problems can be mapped into a classic problem of finding the ground-state of an Ising Hamiltonian. As such, building physical dynamical systems with collective computational ability and distributed parallel processing capability can accelerate the ground-state search. In this talk, we will present a continuous-time dynamical system (CTDS) approach using a coupled network of stochastic phase-transition nano-oscillators (PTNO) where the ground-state solution naturally appears as stable points or attractor states. We will demonstrate the ability of such a network to solve benchmark NP-hard graph partitioning problems with massive speed-up and energy-efficiency when compared to digital CMOS and other contemporary Ising solver approaches.


Bio: Sourav Dutta is currently a Research Assistant Professor in the Electrical Engineering department at University of Notre Dame. He received his Bachelors in Electrical Engineering from Jadavpur University, Kolkata, India, in 2012 and Ph.D. in Electrical and Computer Engineering from Georgia Institute of Technology, Atlanta, USA, in 2017 under Dr. Azad Naeemi. He was a postdoctoral research scholar in Dr. Suman Datta’s group at University of Notre Dame from 2018 - 2020. His current research focuses on emerging device architecture exploration and implementation to enable brain-inspired computing. His work involves experimental characterization, modeling and simulation of ferroelectric and phase-change materials for enabling neural networks and coupled oscillator based neuromorphic hardware.  Previously, he has worked on modeling and simulation of spintronic devices and interconnects for Beyond-CMOS application during his PhD. From May to July 2016, he was a visiting researcher at IMEC, Belgium where he was involved in modeling and simulation of nanoscale plasmonic logic gates, magneto-plasmonics and spin-plasmonics for boolean and non-boolean computation for Beyond-CMOS application. Sourav is the recipient of the 2018 Sigma Xi Best Ph.D. Thesis Award at Georgia Tech and 2009 Kishore Vaigyanik Protsahan Yojana (Young Scientist Encouragement Program) Fellowship in India.


Bio: Suman Datta is the Stinson Professor of Nanotechnology in the Department of Electrical Engineering at the University of Notre Dame, Notre Dame, Indiana, where he directs research in the Nanoelectronic Devices Laboratory. He is also the Director of the SRC/DARPA sponsored Applications and Systems-driven Center for Energy-efficient integrated NanoTechnologies (ASCENT). In addition, he is the Director of the SRC/NSF sponsored Center for Extremely Energy Efficient Collective Electronics (EXCEL). His research involves brain-inspired computing, high performance general-purpose computing and collective state computing with extended CMOS and post CMOS devices. Prof. Datta has co-authored 9 book chapters and over 375 refereed journal and conference papers, and holds 185 patents. His work has received over 24,500 citations (h-index = 81).


This meeting is only available to the JUMP research community, such as Principal Investigators, Postdoc researchers, Students, and Industry/Government liaisons.