SPIDER The Spider Objects
Reduced set selection by L1 Penalizer. a = rss_l1(alg,hyper) generates a rss object, using the l1 norm hyperparameters: child=svm algorithm worked on reoptimize_b=1 recalculate the threshold b0 alpha_cutoff=-1 throw away svs with abs(alpha)optimizer='andre' a.lambda=0 regularizer for selection set to 0 for automatic selection penalize_small=1 min \sum (1/alpha_i) beta_i dont_use_noisy_pts=0 discards noisy points reoptimize=1 model: alpha new alphas for rs-vectors Xsv rs vectors b0 the threshold w2 final value of ||w*-w^2||^2, set to -1 to calculate stats: w2=0 final value of ||w-w*||^2 res=[] results on a separate test set dtst=[] separate test set test_on=0 iterations to test on methods: train constructs a reduced set, returns trained rs-machine test tests new rs-machine on supplied data example: d=gen(toy2d('2circles','l=100')); [r,a]=train(svm({kernel('rbf',1),'C=10000','alpha_cutoff=1e-2'}),d); [r,a2]=train(rss_l1(a,'lambda=1e-2'),d); test(a2,d,loss) author: goekhan bakir, jason weston reference: fast binary and multi-output rss, 2004