New Gradient Techniques for Traces of Functions of Rectangular Matrices and Their Pseudo-Inverses,
Abstract
The generalized inverse is of increasing importance for estimation and optimization in modern systems theory, because it both simplifies many problems and reveals underlying structures of theoretical importance. Optimization, using the gradient of the trace of products of matrix valued functions, including the generalized inverse, are presented in a novel state-space setting. A number of functions of such products of matrices, including general formulas, are derived for the first time to our knowledge. Tables of a number of them are given with some applications in the fields of estimation and optimization. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1977
- Accession Number
- ADA044808
Entities
People
- James S. Pappas.
- Oren N. Dalton