*COMP
*COMP, Matrix
,
Algorithm
, THRESHOLD
,
Val1
, Val2
Compresses a matrix using a specified
algorithm.
Argument Descriptions
-
Matrix
Name of the matrix to compress.
-
Algorithm
Algorithm or method to use:
- SVD --
Singular value decomposition algorithm (default).
- MGS --
Modified Gram-Schmidt algorithm.
- SPARSE --
Compress a sparse matrix based on the threshold value.
-
THRESHOLD
Numerical threshold value used to manage the compression. The default value depends on the method of compression: 1E-7 for SVD; 1E-14 for MGS; 1E-16 for SPARSE.
Val1
and Val2
are additional input used only for the SVD algorithm:
Notes
The SVD and MGS algorithms are only applicable to dense matrices that were created using the *DMAT command. Columns that are linearly dependent on others are removed, leaving the independent or basis vectors. The matrix is resized according to the new size determined by the algorithm.
For the SVD algorithm, the singular value decomposition of an input matrix is a factorization of the form:
Here, the matrix is replaced by the matrix, according to the specified threshold.
The SPARSE compression method is only applicable
to sparse matrices that were created using the *SMAT command.
All terms that have an absolute value below the specified threshold,
relative to the maximum value in the matrix, are removed from the
original matrix. For example, given a sparse matrix having 100
as the largest term and THRESHOLD
=
0.5, all terms having an absolute value below 0.5*100 = 50 are
removed.