Function to optimize the projection index

Share:

Description

This function provides an alternative way to optimize the projection index, by moving along a geodesic path.

Usage

1
caller(start, index, n, bases)

Arguments

start

The Starting Projection for the optimization.

index

The Projection Index function. Typically generated by a call to the pp function.

n

The number of new bases to try at every stage of the optimization. Needs to be an array of the same length as bases. Typically, you either pass a constant vector, or you use a vector with ascending entries, so that you can try more matrices as the optimization proceeds.

bases

The number of new bases desired. Actual number generated may be lesser if optimization stalls.

Details

This function provides an alternative way to optimize the projection index. It moves the index along geodesic paths between randomly generated nearby matrices, in hopes of uncovering peaks of the index function. By experience, one can say that it can often reveal structure missed by Simulated Annealing optimization.

Value

A list of basis matrices, of length bases or shorter (if the optimization stalls).

Author(s)

Mohit Dayal

See Also

Colon