On Tue, Feb 24, 2009 at 07:40:03PM +0100, Michael Hanke wrote: > On Tue, Feb 24, 2009 at 01:35:03PM -0500, Per B. Sederberg wrote: > Memory seems to be fine -- between 15 and 25% of the 12 Gigs. > > > Can you try performing an easy feature selection of say 5-10K > > features, first? One to try might be the new feature stability > > measure I added that looks for the most stable features across runs > > for each label (different from ANOVA): > > > > stab_enet = FeatureSelectionClassifier( > > ENET(lm=1.0,max_steps=500,trace=False,normalize=False), > > SensitivityBasedFeatureSelection( > > CorrStability(), > > FixedNElementTailSelector(5000,mode='select',tail='upper')), > > descr="ENET on 5K best(CorrStability) features") > > Thanks for the code -- will try it now.
That was a lot quicker -- although performs is chance (52%), whenever SMLR got 62% (lm=0.001) and even plain LinearCSVMC got 59%. Although alltogether that is probably also chance ;-) Will play a bit more. Thanks anyway, Michael -- GPG key: 1024D/3144BE0F Michael Hanke http://apsy.gse.uni-magdeburg.de/hanke ICQ: 48230050 _______________________________________________ Pkg-ExpPsy-PyMVPA mailing list [email protected] http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa
