reverting matrix inversion change, was: matrixexp no longer works with a really easy matrix



On 2013-06-19, Barton Willis <willisb at unk.edu> wrote:

> This bug is due to circa January 2013 changes to matrix inversion.

At this point I think we've run into several problems caused by changing
the default matrix inversion method. I'd like to revisit that. I'm not
inclined to go back to the status quo ante but let's see if we can make
some other choice. How about this: use adjoint method if the matrix is
"small", otherwise use something else (code for inverse by LU
decomposition or Gaussian elimination already exist, I guess we could
write something else). What do you think of that?

best

Robert Dodier