[ugrads] Jagadeeswaran Rathinavel's comprehensive exam, 8 AM, Tuesday, October 17, in RE 122

Fred Hickernell hickernell at iit.edu
Thu Oct 12 11:07:53 CDT 2017


Dear colleagues and students,

Please join us for an upcoming comprehensive exam.  

(Students, do you wonder what constitute’s math research, or what a PhD student has to do, or how they get grilled by their PhD committee?  Whether it is out of curiosity, to provide moral support for Jagadees, or to prepare yourself for what is coming up for you, please join us.)

Candidate: Jagadeeswaran Rathinavel

Title: Automatic Bayesian Cubature

Abstract:  Automatic cubatures are any class of algorithms which offer integration result within the user specified error threshold by adjusting the complexity, and also increasing the number of data samples used. Bayesian cubature uses the prior information about the integrand function to improve the accuracy. Bayesian cubature provides faster convergence rates with confidence intervals, but requires time-consuming calculations. So Bayesian cubature methods without any special techniques are usually computational intensive. The goal of this work is to develop a Fast and Guaranteed Automatic Integration algorithm based on Bayesian concepts. We use a combination of low discrepancy points sets and a nicer kernel to optimize the computational cost.  We will obtain an error bound using posterior error distribution. We formulate a general theory called Fast Transform kernel so that this idea can be used with a smooth kernel and an appropriate low discrepancy  point sets. We demonstrate this method a special shift invariant kernel and Rank-1 Lattice points. This is my thesis work with Prof Fred. Hickernell.

Place: RE 122

Time: 8 AM, Tuesday, October 17, 2017

Best regards,
Fred


Fred J. Hickernell
Director, Center for Interdisciplinary Scientific Computation
Professor, Department of Applied Mathematics
Illinois Institute of Technology
Pritzker Science Ctr Rm 106, 3105 S Dearborn St, Chicago, IL 60616
hickernell at iit.edu <mailto:hickernell at iit.edu>, www.iit.edu/~hickernell <http://www.iit.edu/~hickernell>
Office: +1 312 567 8983   Cell: +1 630 696 8124   Skype: fjhickernell

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://math.iit.edu/pipermail/ugrads/attachments/20171012/e4be1724/attachment.html>


More information about the ugrads mailing list