Optimization and Control Seminar, Spring 2008

Thursday, March 6, 2008, 11:00-12:00 SOND 406
Higher order corrector-predictor infeasible interior point method for sufficient linear complementarity problems
- Mr. Adrian Vancea
Abstract: I will present a corrector-predictor method for solving sufficient linear complementarity problems (LCP) with an infeasible starting point. The method generates a sequence of iterates in an $l_2$-neighborhood of the infeasible central path of the LCP without depending on the handicap $\kappa$ of the problem, so it may be used for any sufficient linear complementarity problem. Our algorithm has the best known iteration complexity if the starting point is feasible or close to be feasible and also if our starting point is "big enough". Moreover, our algorithm is superlinearly convergent even for degenerate problems. More precisely, by using a predictor with order $m>1$, we show that the duality gap converges to zero with Q-order $m+1$ in the nondegenerate case, and with Q-order $(m+1)/2$ in the degenerate case. Some numerical results are given to demonstrate this method's effectiveness. Note: This is a joint work with: Dr. Potra, C. Petra, F. Gurtuna and O. Schevchenko
 
Thursday, March 13, 2008, 12:00-1:00 pm SOND 406
Positive Invariance of Constrained Affine Dynamics and Its Applications to Hybrid Systems and Safety Analysis
- Dr. Jinglai Shen
Abstract: This talk is the first part of a series of two talks on global long-time behaviors of constrained affine dynamics, motivated by recent work on piecewise affine systems (PASs) and reachability analysis of linear/affine dynamics. We will start with a brief review of local and finite-time switching properties of (Lispchitz continuous) PASs, for example, non-Zenoness and simple switching behavior, and discuss their applications to dynamical complementarity systems. We then focus on global long-time switching behaviors of PASs. We show that the positively invariant set associated with each affine dynamics plays an important role in characterization of various global switching dynamics of PASs. We will discuss certain properties of positively invariant sets and their implications to PASs and reachability analysis.
 
Thursday, March 27, 2008, 12:00-1:00 pm SOND 406
Some inertia theorems in Euclidean Jordan algebras
- Dr. Jiyuan Tao of Loyola College
Abstract: In this talk we present some inertia theorems in Euclidean Jordan algebras. First, based on the continuity of eigenvalues, we give an alternate proof of Kaneyuki's generalization of Sylvester's law of inertia in simple Euclidean Jordan algebras. As a consequence, we show that the cone spectrum of any quadratic representation with respect to a symmetric cone is finite. Second, we present Ostrowski-Schneider type inertia results in Euclidean Jordan algebras. In particular, we relate the inertias of objects $a$ and $x$ in a Euclidean Jordan algebra when $L_a(x)>0$ or $S_a(x)>0$, where $L_a$ and $S_a$ denote Lyapunov and Stein transformations respectively. This is a joint work with M.S. Gowda and M. Moldovan.
 
Thursday, April 3, 2008, 12:00-1:00 pm SOND 406
Some inertia theorems in Euclidean Jordan algebras
- Dr. Jiyuan Tao of Loyola College
Abstract: This talk will be continuity of the last week talk, we present Ostrowski-Schneider type inertia results in Euclidean Jordan algebras. In particular, we relate the inertias of objects $a$ and $x$ in a Euclidean Jordan algebra when $L_a(x)>0$ or $S_a(x)>0$, where $L_a$ and $S_a$ denote Lyapunov and Stein transformations respectively.
 
Thursday, April 10, 2008, 12:00-1:00 pm SOND 406
Duality of ellipsoidal approximations via semi-infinite programming
- Ms. Filiz Gurtuna of UMBC
Abstract: Finding the minimum volume ellipsoid containing a convex body and the maximum volume ellipsoid contained in a convex body are very important problems having applications in optimization, statistics, data mining and control theory, among others. As the numerical computation of these ellipsoids is needed in all the applications mentioned, the study of the dual problems becomes crucial. In this work, we develop duality of the minimum volume circumscribed ellipsoid and the maximum volume inscribed ellipsoid problems. We present a unified treatment of both problems using convex semi--infinite programming. We establish the known duality relationship between the minimum volume circumscribed ellipsoid problem and the optimal experimental design problem in statistics. The duality results are obtained utilizing (in fact, developing) convex duality for semi--infinite programming in a functional analysis setting.
 
 

Past Optimization and Control Seminars