ELI5: Multi-dimensional Integration (3/7/2014 Hudson 208 @ 11:45AM)

Justin Winokur
Friday, March 7, 2014 - 11:45am
Hudson 208


Numerical integration (or quadrature) is a pervasive problem in data analysis and in computational research. In a low-dimensional setting, the problem is small, however, work required scales exponentially with number of dimensions and becomes intractable very quickly.

In this casual tutorial, Justin will demonstrate and explain a technique called Smolyak Quadrature to ameliorate some of the issues with traditional quadratures. Furthermore, he will demonstrate adaptive methods suitable to a wide class of problems.

These techniques may prove to be a useful arrow in your quiver of computational tools, even if you do not require them immediately. No prior knowledge is needed or expected.




Justin Winokur is a 4th year PhD candidate at Duke. He holds a B.S. in physics from Carnegie Mellon University and an M.S.E. in mechanical engineering from The Johns Hopkins University where he attended before transferring to Duke.