Re: [Gretl-devel] testcases and pet projects
On Sun, 5 Jul 2009, Klein, Christoph wrote:
Hi,
I've attached a badly specified mle model which provokes the error. I
don't know whether one can find a well specified model which has these
problems, too. The script is named test_bfgs.inp.
Thanks. This is very helpful. The problem lies with the fact that we don't
gauge the steplength properly at the first iteration. In practice, the
algorithm starts by jumping to a wrong point, and never comes back. To
demonstrate this is the case, if you artificially force a shorter first
step by something like
--- lib/src/gretl_bfgs.c.~1.7.~ 2009-07-06 11:07:30.000000000 +0200
+++ lib/src/gretl_bfgs.c 2009-07-06 11:24:04.000000000 +0200
@@ -500,7 +500,7 @@
for (j=0; j<i; j++) {
H[i][j] = 0.0;
}
- H[i][i] = 1.0;
+ H[i][i] = 0.01;
}
}
BFGS converges alright. I'll see what I can do.
I've attached two more patches. [...]
Wow, thanks, this is great! As Allin said, though, I'm afraid it's just
sensible if this waits after 1.8.2.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti