def sample_combinations(samples, sample_labels, per_class, classifier, alg_args):
n_features = shape(samples)[1];
features = tuple(range(n_features));
ultimate = (-inf, features),;
per_k = 4;
subsets = [];
for k in range(1, n_features):
combs = [comb for comb in combinations(features, k)];
subsets += native_sample(combs, min([per_k, len(combs)]));
candidates = tuple((merit(samples, sample_labels, comb), comb) for comb in subsets);
return candidates;
def stochastic_selector(samples, frontier, successors, beam_width, classifier):
successors = tuple(native_sample(successors, min((beam_width, len(successors)))));
return tuple((merit(samples, sample_labels, s), s) for s in successors);
