Green's functions and large matrix systems: electronics and multiphysics design, networked systems, and recommendation engines.

Department of Electrical Engineering

University of Washington

This talk summarizes approaches to physics-based compression and scaling of large Green's function-based matrix systems. Examples from electronic design automation, electromagnetic simulation, and microfluidic multiphysics systems are discussed. As scalable and parallelized physics-solvers have become prevalent, the focus has moved from sub-system analysis to overall system design, bringing with it a new set of challenges associate with high-dimensional regression, optimization, incremental simulation, and design under uncertainty. These will be discussed with examples of RFID system design and approaches to wireless power transmission. Combining the geometry of physical systems and the topology of abstractions from circuit and network theory has also moved the challenges more towards graph-theoretic issues. New emerging areas in the application of Green's function-like techniques towards "big data" problems in internet and social network search paradigms and cloud-based distributed secure computation will be summarized.