Orange Forum • View topic - Vizrank does not work the same

Vizrank does not work the same

Report bugs (or imagined bugs).
(Archived/read-only, please use our ticketing system for reporting bugs and their discussion.)
Forum rules
Archived/read-only, please use our ticketing system for reporting bugs and their discussion.

Vizrank does not work the same

Postby lareooreal_2 » Fri Sep 11, 2009 9:25

Dear all!

I am stuck with the fact that the Vizrank implementations
do not work the same depending in which widget they are
implemented. I am using the Vizranks within
Linear Projection, Radviz and Polyviz and it happens that
using the same settings


- the list of combined attributes is not prioritized in the same
manner. A set of topcandidates in one method is largely differing
from the set found in the other method, despite I am using the same
settings (RelieF, number et etc)

- if a set of combined attributes is present in two independent
Vizrank runs of two widgets then it happens that the classification
accuracy computed differs. I.E. the ROC performance or others
are not computed reproducibly for the class speparation problem


How can I avoid this ?


A minor annoyance in Vizrank is that there is no possiblity
to set a boundary condition to set the miniumum number
of representations. In a matrix containing non availables
it happens that the selected atrribute set have only
minor representation (but good classification power)

Cheers Lareooral

Return to Bugs



cron