MATPINV Matrix PseudoInverse
The MATPINV Block outputs the inverse of a non square input matrix using the SVD (Singular Value Decomposition) operation. If the SVD of A is:
A=USV'
The pseudoinverse X of A is given by:
X=VS"U' where S"(i,j)=1/S(i,j) (if S(i,j) =0), U' and V are respectively the transpose of U and V'.
and we have A*X*A=A and X*A*X=X. Both A*X and X*A are Hermitian matrices. A warning message is printed if the input is badly scaled or nearly singular.
When the input is an M-by-N matrix, the output is an N-by-M matrix. The equivalent function of this block in Scilab is "pinv".
- Datatype(1=real double 2=Complex)
It indicates the type of the output. It support only the two types double (1) and complex (2). If we input another entry in this label Scicos will print the message "Datatype is not supported".
Properties : Type 'vec' of size 1.
- always active: no
- direct-feedthrough: yes
- zero-crossing: no
- mode: no
- regular inputs:
- port 1 : size [-1,-2] / type 1
- regular outputs:
- port 1 : size [-2,-1] / type 1
- number/sizes of activation inputs: 0
- number/sizes of activation outputs: 0
- continuous-time state: no
- discrete-time state: no
- object discrete-time state: no
- name of computational function: mat_pinv
- SCI/macros/scicos_blocks/MatrixOp/MATPINV.sci [view code]
Fady NASSIF INRIA
