Distinguished Seminar Series in Computational Science and Engineering
Thursday, April 27th 2017 | 12:00 PM* | 37-212

A new tensor framework - theory and applications
Misha Kilmer
Professor of Mathematics, Adjunct Professor of Computer Science
Tufts University

Tensors can be instrumental in revealing latent correlations residing in high dimensional spaces. Despite their applicability to a broad range of applications in machine learning, speech recognition, and imaging, inconsistencies between tensor and matrix algebra have been complicating their broader utility. Researchers seeking to overcome those discrepancies have introduced several different candidate extensions, each introducing unique advantages and challenges. In this talk, we review some of the common tensor definitions, discuss their limitations, and introduce our tensor product framework which permits the elegant extension of linear algebraic concepts and algorithms to tensors. Following introduction of fundamental tensor operations, we discuss in further depth tensor decompositions and in particular the tensor SVD (t-SVD) and its randomized variant, which can be computed efficiently in parallel. We present details of the t-SVD, theoretical results, and provide numerical results that show the promise of our approach for compression and analysis of operators and datasets, highlighting examples as facial recognition, video completion and model reduction.

* Lunch provided at 11:45

Info Session: Ardavan Oskooi, Sc.D.
Founder & CEO, Simpetus
May 4th | 12-1PM* | 2-105

Leveraging Advances in Computational Electrodynamics to Enable New Kinds of Nanophotonic Devices
Ardavan Oskooi

Advances in computational electrodynamics have the potential to enable fundamentally new kinds of nanophotonic devices which are based principally on complex, non-analytical wave-interference effects. Powerful, flexible, open-source software tools have now been made available for use in large-scale, parallel computations to model the interaction of light with practically any kind of material in any arbitrary geometry. These recent developments in computational capability make possible the investigation of various emergent structures, materials, and physical phenomena that were previously beyond the reach of pencil and paper analytical methods as well as less versatile and even less accessible commercial software tools. I will demonstrate how such advances in finite-difference time-domain (FDTD) methods for computational electromagnetics via the open-source software package MEEP can lead to new designs for light trapping in thin-film silicon solar cells as well as light extraction from organic light-emitting diodes (OLEDs). I will then describe efforts by our startup to leverage scalable, cluster computing in the public cloud for large-scale device design and also talk briefly about the journey from academic research to technology entrepreneurship.

Ardavan Oskooi is the Founder and CEO of Simpetus, a San Francisco-based startup with a mission to propel simulations to the forefront of research and development in electromagnetics. Simpetus is a reference to our vision for simulations being an impetus for new discoveries and technologies. Ardavan received his Sc.D. from MIT where he worked with Professors Steven G. Johnson and John D. Joannopoulos to develop MEEP (thesis: Computation & Design for Nanophotonics). Ardavan has published 13 first-author articles in peer-reviewed journals and the book Advances in FDTD Computational Electrodynamics: Photonics and Nanotechnology with Professors Allen Taflove of Northwestern University and Steven G. Johnson. He has a masters in Computation for Design and Optimization from MIT and completed his undergraduate studies, with honors, in Engineering Science at the University of Toronto. Prior to launching Simpetus, Ardavan worked as a postdoctoral researcher with Professors Susumu Noda at Kyoto University and Stephen R. Forrest at the University of Michigan on leveraging computational electromagnetics to push the frontier of optoelectronic device design.

*Lunch Provided