Unexpected result calculating the determinant of a singular matrix (42S)

[Valentin Albillo]

(10-21-2019 02:30 AM)Dave Britten Wrote:  Okay, I kind of wondered if there were some tricks like that going on. I remember seeing mention of LU decompositions in the Advantage Module manual, so I'm assuming that gives a similar result. The 48G, on the other hand, does return a determinant of 0 for that particular matrix, as does the 3x3 linear system solver from the 32SII manual.

The 48G "cheats": it detects that all elements are integer and forces the result (computed using LU as well) to be the nearest integer, which succeeds sometimes and fails some others.

Try the 7x7 matrix I give as an example in the linked paper in the 48G and see if you get 1 as the determinant. You won't, the cheat doesn't work.

V.