Dear Root-ers, I have a problem with the Invert operation in TMatrix. When trying to invert a 250 x 250 matrix with only diagonal elements of size 10000, I get a floating-point exception. Trying inversion of a 5 x5 matrix works ok. The remedy is to first normalize the matrix properly, invert and transform back. I did not look in the code but assume this is not done. Could the ROOT team please add something to the code that will take care of this: (pseudo code) 1. b(i,j) = a(i,j)/sqrt(a(i,i)*a(j,j)) 2. b_inv = b.Invert() 3. a_inv(i,j) = b_inv(i,j)/sqrt(a(i,i)*a(j,j)) Eddy Offermann
This archive was generated by hypermail 2b29 : Tue Jan 02 2001 - 11:50:24 MET