Speaker: Dr. Arijit Raychowdhury, GeorgiaTech University
Collective dynamical systems offer unique opportunities for computing by harnessing the complex interactions of simple elements such as oscillators or spike generators. This is possible, when such dynamics can be programmed, controlled, and observed. In this talk, I will present some of our work where we are exploring the time-evolution of both deterministic and stochastic dynamical systems in both CMOS and post-CMOS computing substrates. I will show applications of such systems in solving inverse problems, distributed optimizations (convex and combinatorial) and machine learning. In particular, I will discuss our recent work that connects dynamics and algebraic graph theory. Finally I will talk about implementation of such dynamics in mixed-signal CMOS, including a recent demonstration of reinforcement learning for energy-constrained edge devices. I will conclude with a brief discussion of the opportunities, potentials and challenges in realizing such computational systems.