Deep Learning the Properties of Metamaterials
Document Type
Presentation Abstract
Presentation Date
11-28-2022
Abstract
Unlike conventional materials, metamaterials derive their properties primarily from their structure rather than their bulk construction materials. With a carefully-chosen structure, electromagnetic metamaterials have been shown to exhibit exotic properties that are not achievable with conventional materials, and now underpin many technologies. In principle even more exotic and useful properties are achievable, but the modeling and design of advanced metamaterials is challenging, and a major bottleneck to continued progress. In this talk I discuss the challenges of modeling and designing advanced metamaterials, and how recent advances in deep learning – a branch of machine learning - have shown the potential to overcome some of these challenges. In particular, I discuss recent deep learning methods – some developed by myself with collaborators at Duke University - that can dramatically accelerate both the modeling and design of complex metamaterials. In principle these methods can also be applied to many other natural systems, accelerating scientific progress and technological development. I close by discussing some open challenges at the intersection of machine learning and scientific computing.
Recommended Citation
Malof, Jordan M., "Deep Learning the Properties of Metamaterials" (2022). Colloquia of the Department of Mathematical Sciences. 643.
https://scholarworks.umt.edu/mathcolloquia/643
Additional Details
November 28, 2022 at 3:00 p.m. Math 103