From the supposedly known inverse of B,
a certain submatrix of A, calculates the inverse of another submatrix,
B', in which one (only) column is different. Matrix B
is made of the V columns of A, and matrix B'
of the W columns. Exactly one element of V and W
must be different.
The advantage of this procedure is that it is simply
B'−1 = E B−1,
thus avoiding its direct inversion (with E a simple,
quasi-identity matrix).
The procedure is used in Linear Programming, typically "linked"
to this algorithm, but is general.
If 'random', the matrix elements are generated in [1, 10].
(Caution: no proviso is included to guarantee
that any of the matrices is non-singular.)
|
