Bounding Ellipsoid Optimization MIT Media Laboratory
		Bounding ellipsoids approximate complex polytopes with well behaved and easy to manipulate algebraic representations. Many such polytopes emerge from constraint-based formulations of phenomena in engineering and science where linear models are accommodated with assumptions on worst case deviations. This short presentation derives a general expression for bounding ellipsoid polytopic supersets, discusses select optimization strategies, sequential convergence properties and potential for reducing computational complexity and power consumption. Sample applications are presented along with a discussion of parallels with Kalman filters and Support Vector Machines. Basic understanding of linear algebra is assumed. [poster]
Date:		January 23rd, 2008
Time:		1-5 pm
Where:		E15-239 (Roth room)
Instructor:		Dale Joachim
Contact:		email: joachimd(at)media.mit.edu tel: 617 452 5635 office: E15-320C