ml_compress_ldl
LDL^T row/column compression.
Contents
Syntax
[Z, Y] = ml_compress_ldl(Z, Y, tol, type)
Description
Computes a column or row compression of the matrices Z and Y using an eigenvalue decomposition. Usually used to approximate the products Z'*Y*Z (row compression) or Z*Y*Z' (column compression) via low-rank factors.
Input
- Z - matrix of dimensions n x m (column compression) or m x n (row compression)
- Y - symmetric matrix of dimensions m x m
- tol - nonnegative scalar, tolerance multiplied with the largest singular value to determine the rank of the approximation
- type - string, switching between row and column compression 'c', 'col', 'column' - for column compression 'r', 'row' - for row compression
Output
- Z - matrix of dimensions n x r in case of column compression and r x n in case of row compression
- Y - matrix of dimensions r x r
See Also