Time: 11am-12:15
Place: CS 220 (new CS building, seminar room, 2nd floor)
Guest speaker/problem poser: Prof. Andrew Mullhaupt
AMS faculty in Quantitative Finance
Come hear about cool algorithmic problems in finance!
Preview:
"I propose to present a particular problem in portfolio selection and the
associated convex optimization which underlies the conventional approach.
Then we find inequalities in the KKT system which characterize the unique
solution as a sort of art gallery problem, which leads to a remarkably fast
O(n) algorithm, which is essentially exact. This algorithm is quite clear.
Then we generalize the problem to a one which has wider applicability and
derive a dynamic programming algorithm which is at worst O(n log n) and
possibly could be sharpened if a closer analysis is done depending on the
search structure chosen...
Finally, the entire point is that these algorithms consider the case of a
one-dimensional portfolio space. A very important case in applications, but
there is significant interest in the generalization to multiple dimensions.
Time permitting, the relationship of the problem to submodularity (i.e.
Monge / convex quadrangle inequalities) and other algorithm accelerations
can be discussed."
Short bio: Andrew Mullhaupt's original research interests involved partial
differential equations and linear algebra. He has worked on Wall Street for
over 20 years designing mathematically based trading systems, first at
Morgan Stanley, then at Renaissance Technologies, and most recently at S A
C Capital where he was Director of Research for the SAC's Meridian Fund.
http://pw1.netcom.com/~amullhau/