A problem in EigenvalueDecomposition
- Subject: A problem in EigenvalueDecomposition
- From: Saikat DebRoy <firstname.lastname@example.org>
- Date: 23 Jul 1999 08:55:32 -0500
I am not sure if I should call it a bug. In Jama.EigenvalueDecomposition,
after the while loop in tql2() in lines 200-205, we should have m <
n. But we may end up with m == n if d[l] or e[l] is a Double.NaN
(comparison with a Double.NaN always return false).
I think you should throw an exception at the end of the loop if m ==
n. Right now, an ArrayIndexOutOfBounds exception is thrown at line 234
- but it is not very informative to the user.
One challenge with converting the FORTRAN routines in LINPACK, EISPACK
etc to Java in a would be to take into consideration the presence of
the three extra floating point numbers in Java - Double.POSITIVE_INFINITY,
Double.NEGATIVE_INFINITY, and Double.NaN.
Date Index |
Thread Index |
Problems or questions? Contact email@example.com