| Paper: | MLSP-P6.7 | ||
| Session: | Learning Theory and Models | ||
| Time: | Thursday, May 20, 15:30 - 17:30 | ||
| Presentation: | Poster | ||
| Topic: | Machine Learning for Signal Processing: Learning Theory and Modeling | ||
| Title: | LINE SEARCH AND GRADIENT METHOD FOR SOLVING CONSTRAINED OPTIMIZATION PROBLEMS | ||
| Authors: | Mohammed Hasan; University of Minnesota, Duluth | ||
| Abstract: | Optimization over linear and orthogonal (or unitary) constraints arises in many applications in eigenvalue regularization, control theory, and signal processing. For example, optimization of symmetric Rayleigh quotient over the unit sphere yields the minimum and maximum eigenvalue of a symmetric matrix. In this paper, many problems involving minimum subspace computation, minor and principal subspace tracking, adaptive subspace computation, computing the first $r$ dominant eigenpairs, canonical correlation analysis, reduced rank Wiener filtering will be solved using similar framework.The main features of these algorithms are 1) they are computationally efficient in that they are matrix inverse free methods, and 2) they are based on gradient descent adaptation with exact or approximate line search. | ||
| Back | |||
Home -||-
Organizing Committee -||-
Technical Committee -||-
Technical Program -||-
Plenaries
Paper Submission -||-
Special Sessions -||-
ITT -||-
Paper Review -||-
Exhibits -||-
Tutorials
Information -||-
Registration -||-
Travel Insurance -||-
Housing -||-
Workshops