ECE SEMINAR: Deep Learning Models for the Design of Metamaterials

Metamaterials are materials that derive their properties primarily from their structure rather than the material from which they are composed. Furthermore, with a carefully-chosen structure, metamaterials have been shown to exhibit powerful electromagnetic properties that are not achievable with conventional materials, and now underpin many technologies. In principle far more powerful properties are achievable however, identifying the structures needed to yield a particular set of EM properties - i.e., metamaterial design - is an open and challenging problem, and a major bottleneck to continued progress. Recently however advances in deep learning have shown tremendous potential to overcome many of the challenges associated with metamaterial design, and enable the design of metamaterials with substantially more powerful properties. In this talk I will discuss contemporary challenges with the design of metamaterials, and recent advances using deep learning to overcome these challenges, including work that my collaborators and I recently published at NeurIps 2020.

Zoom: https://duke.zoom.us/j/91478962111?pwd=QUZwY0Q3ZDB2NWZxVnJKS3g1WU0zZz09;
PASSCODE: 347406