Orange.classification.svm.SVMLearner __init__ parameter weight produces wrong results

[damonmaria]

No matter what values I pass in for Orange.classification.svm.SVMLearner weight I get the same confusion matrix out.

I have 5 classes. And have tried weight being proportional to C (as per LibSVM) and so passed:
weight=[1.0, 1.0, 1.0, 1.0, 1.0]

And weight being absolute (same as the C value I use):
weight=[8.0, 8.0, 8.0, 8.0, 8.0]

Those values and all other values I've tried produce the same confusion matrix which is wildly different from the confusion matrix I get without passing weight values. So I don't think this is working.

I have tried to trace the code but I end up at set_weights in what I think is a compiled C .so file.

[ales-erjavec]

The code in orange which passes the weights to libsvm was completely broken (the weight parameter was also documented wrong). I made some fixes in 2db50d0 and 2db50d0, but it still needs to be tested.