Blatt 7, aufgabe 3

5 views
Skip to first unread message

Angelina Duka

unread,
Jun 4, 2011, 7:43:47 AM6/4/11
to ml-tu
Hallo,

ich habe bekomme immer wieder ein Fehler

??? Error using ==> chol
Matrix must be positive definite.

Error in ==> sheet07_2>pr_loqo2 at 50
R = chol(H_x); H_Ac = R \ ([A; c_x'] / R)';

Error in ==> sheet07_2>oneclass at 121
[alpha,y] = pr_loqo2(c, H, A, b, l, C);

Daraufhin habe ich die gegebene Matrix überprüft.
Sie hat 236 negative Eigenwerte.

Der Kern sollte positiv definit sein.

Gibt es einen Fehler in den Daten?

Felix Brockherde

unread,
Jun 4, 2011, 8:26:40 AM6/4/11
to ml...@googlegroups.com
Hi
Wir hatten das gleiche Problem. Du kannst zur Matrix eye(n)*1e-10
addieren und es funktioniert. Wir hatten im Endeffekt einen Vektor
falsch übergeben und danach lief es auch ohne Regularisierung.

Viele Grüße, Felix

priska herger

unread,
Jun 4, 2011, 12:09:45 PM6/4/11
to ml...@googlegroups.com
hey e"one,

when running pr_loqo2 with the specified args, i get the following error

??? Error using ==> mrdivide
Matrix dimensions must agree.

Error in ==> sheet07>pr_loqo2 at 50

R = chol(H_x); H_Ac = R \ ([A; c_x'] / R)';

despite sticking to the dimensionality of the input args as specified.

did any of you have a similar problem?
and if so, any hints?

salut,
priska

Felix Brockherde

unread,
Jun 4, 2011, 1:57:26 PM6/4/11
to ml...@googlegroups.com
Are you sure the dimensionality is correct? It's working for us. N-row vector means 1xN-vector.
You can always put more line breaks and introduce temp variables to see exactly what goes wrong.
Reply all
Reply to author
Forward
0 new messages