subspacea - Angles between subspaces

Version 4.6 (11,9 KB) von Andrew Knyazev
Angles between subspaces and canonical correlations in general scalar products.
3,9K Downloads
Aktualisiert 20 Dez 2021

Lizenz anzeigen

subspacea(F,G,A) Finds all min(size(orth(F),2),size(orth(G),2)) principal angles between two subspaces spanned by the columns of matrices F and G in the A-based scalar product x'*A*y, where A is Hermitian and positive definite. COS of principal angles is called canonical correlations in statistics.
The input A can be provided as a matrix as well as a string of a function name, e.g., funcA giving the product funcA(X) = A*X.
[theta,U,V] = subspacea(F,G,A) also computes left and right principal (canonical) vectors - columns of U and V, respectively.
If F and G are vectors of unit length and A=I, the angle is ACOS(F'*G) in exact arithmetic. If A is not provided as a third argument, than A=I and the function gives the same largest angle as SUBSPACE.m by Andrew Knyazev. MATLAB's SUBSPACE.m function is still badly designed and fails to compute some angles accurately.
The algorithm is described in A. V. Knyazev and M. E. Argentati, Principal Angles between Subspaces in an A-Based Scalar Product: Algorithms and Perturbation Estimates. SIAM J. Scientific Computing, 23 (2002), no. 6, 2009-2041. http://dx.doi.org/10.1137/S1064827500377332

Zitieren als

Andrew Knyazev (2024). subspacea - Angles between subspaces (https://www.mathworks.com/matlabcentral/fileexchange/55-subspacea-angles-between-subspaces), MATLAB Central File Exchange. Abgerufen .

Knyazev, Andrew V., and Merico E. Argentati. “Principal Angles between Subspaces in an A-Based Scalar Product: Algorithms and Perturbation Estimates.” SIAM Journal on Scientific Computing, vol. 23, no. 6, Society for Industrial & Applied Mathematics (SIAM), Jan. 2002, pp. 2008–40, doi:10.1137/s1064827500377332.

Mehrere Stile anzeigen
Kompatibilität der MATLAB-Version
Erstellt mit R2021a
Kompatibel mit allen Versionen
Plattform-Kompatibilität
Windows macOS Linux
Kategorien
Mehr zu Dimensionality Reduction and Feature Extraction finden Sie in Help Center und MATLAB Answers
Tags Tags hinzufügen
Quellenangaben

Inspiriert von: subspace.m, ortha.m

Inspiriert: subspace.m

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!
Version Veröffentlicht Versionshinweise
4.6

Fix error handling, code beautification.

1.2

Updated description.
added a conversion to a toolbox

1.1.0.0

License update to free software (BSD). Comments update.

1.0.0.0

minor update for MATLAB R14