The sign matrix and the separation of matrix eigenvalues
Citations
217 citations
Cites background or methods from "The sign matrix and the separation ..."
...A globally, asymptotically quadratically convergent iteration to compute the sign-function of B is Bi+1 = (Bi + B 1 i )=2 [108, 168, 190]; this is simply Newton's method applied to B(2) = I , and can also be seen to equivalent to repeated squaring (the power method) of the Cayley transform of B....
[...]
...One such function f is the sign function [13, 108, 120, 135, 168, 190] which maps points with positive real part to +1 and those with negative real part to 1; adding 1 to this function then maps eigenvalues in the right half plane to 2 and in the left plane to 0, as desired....
[...]
...By working on a shifted and squared real matrix, one can divide along lines at an angle of =4 and retain real arithmetic [13, 108, 190]....
[...]
212 citations
210 citations
145 citations
145 citations
References
7,422 citations
430 citations
267 citations
229 citations