I encountered similar problems.
Weighting libsvm inputs under sklearn seems unsafe at any speed...
Amusingly enough, I think the pretty demo at [1] demonstrates the
problem. Notice how the white samples are really dominant... but the
decision boundary is consistent with uniform weights.
If you change the script slightly, its obvious changing the weight
ratio does not change the displayed decision boundary, so it is not a
problem of merely scaling or interaction with C.
This may be related to the fact that sample weighting in libsvm itself
does not appear to be a stable core feature; [2] implies that a
special version of libsvm is needed. Or is that already taken into
account?
[1] http://scikit-learn.org/stable/auto_examples/svm/plot_weighted_samples.html
[2] http://www.csie.ntu.edu.tw/~cjlin/libsvmtools/#weights_for_data_instances
--
Daniel Vainsencher
------------------------------------------------------------------------------
See everything from the browser to the database with AppDynamics
Get end-to-end visibility with application monitoring from AppDynamics
Isolate bottlenecks and diagnose root cause in seconds.
Start your free trial of AppDynamics Pro today!
http://pubads.g.doubleclick.net/gampad/clk?id=48808831&iu=/4140/ostg.clktrk
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
