leigopt : Large-Scale Optimization of Eigenvalues
Here you can access the Matlab routines for large scale eigenvalue and singular value optimization. The main routines to call are as follows:
leigopt_max (to maximize the Jth largest eigenvalue)- leigopt_min (to minimize the Jth largest eigenvalue)
- lsvdminopt_min_general (to minimize the Jth smallest singular value)
Download the routines
The scope of the routines (what can be solved with these routines)
How to call the routines
Sample calls
The routines here are based on the subspace procedures described in the following paper:
A subspace method for large-scale eigenvalue optimization (together with F. Kangal, K. Meerbergen and Wim Michiels), submitted to SIMAX.
The figure on the top is an illustration of the subspace procedure to minimize a largest eigenvalue. The solid curve is the eigenvalue function, the dashed and dotted curves are reduced eigenvalue functions formed by the subspace procedure.