Title:  Computing with Dynamical Systems: Bridging Device Technologies with Algorithms

Committee:

Dr. Arijit Raychowdhury, ECE, Chair , Advisor

Dr. Justin Romberg, ECE

Dr. Saibal Mukhopadhyay, ECE

Dr. Magnus Egerstedt, ECE

Dr. Suman Datta, Univ of Notre Dame

Abstract:

The objective of the proposed research is to create alternative computing models, unlike (discrete) sequential Turing machine/Von Neumann style models, which utilize the unique behavior of novel beyond-CMOS devices. This work focusses on circuits (mainly coupled oscillators) composed of hysteretic insulator-metal-transition (IMT) devices. Such circuits can have complex dynamics at the network level which can be used to emulate higher level functions than merely switches, logic gates or arithmetic operations. Network level circuits of IMT devices and their dynamics are studied across 3 levels - algorithms, dynamical systems and circuits/devices. Some computing models are proposed where the dynamics calculates and encodes solutions like analog subtraction, graph coloring and matrix eigenvectors in the steady state phases of coupled oscillator networks. The oscillator dynamics are equivalent to some spiking neuron models, and as such, this also forms a model of spike-time encoding of solutions.