Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
Drexel University or The Math Forum.



Q: on matrix inverse by partitioning
Posted:
Jul 15, 1996 9:56 PM


'numerical recipes' lists the following partitioning method for finding matrix inverses. given:
A = [ P  Q ] and Inv(A) = [ P*  Q* ] [ R  S ] [ R*  S* ]
where P and S are square not necessarily of the same size:
P* = Inv(P  Q.Inv(S).R) Q* = Inv(P  Q.Inv(S).R).Q.Inv(S) R* = Inv(S).R.Inv(P  Q.Inv(S).R) S* = Inv(S) + Inv(S).R.(P  Q.Inv(S).R).Q.Inv(S)
but states nothing about the case when A is invertible but S is not for example. does anyone know the set of conditions that A must satisfy for this algorithm to be valid?
thanks for the info,
alan



