def test_voting():
base = ctransforms(milk.supervised.svm.svm_raw(C=2.,kernel=milk.supervised.svm.rbf_kernel(2.**-3)),milk.supervised.svm.svm_binary())
base = milk.supervised.multi.one_against_rest(base)
features,labels = milksets.wine.load()
gfeatures, glabels = group(features, labels, 3)
learner = milk.supervised.grouped.voting_classifier(base)
learner.train(gfeatures, glabels)
model = learner.train(gfeatures, glabels)
assert ([model.apply(f) for f in gfeatures] == np.array(glabels)).mean() > .8
def test_ecoc_probability():
features,labels = load()
features = features[labels < 5]
labels = labels[labels < 5]
raw = svm.svm_raw(kernel=svm.dot_kernel(), C=1.)
base = ctransforms(raw, svm.svm_sigmoidal_correction())
learner = ecoc_learner(base, probability=True)
model = learner.train(features[::2], labels[::2])
results = list(map(model.apply, features[1::2]))
results = np.array(results)
assert results.shape[1] == len(set(labels))
assert np.mean(results.argmax(1) == labels[1::2]) > .5
import numpy as np
import random
import milk.supervised.svm
import milk.supervised.multi
from milk.supervised.classifier import ctransforms
from .fast_classifier import fast_classifier
import milksets.wine
features,labels = milksets.wine.load()
A = np.arange(len(features))
random.seed(9876543210)
random.shuffle(A)
features = features[A]
labels = labels[A]
labelset = set(labels)
base = ctransforms(milk.supervised.svm.svm_raw(C=2.,kernel=milk.supervised.svm.rbf_kernel(2.**-3)),milk.supervised.svm.svm_binary())
def test_one_against_rest():
M = milk.supervised.multi.one_against_rest(base)
M = M.train(features[:100,:],labels[:100])
tlabels = [M.apply(f) for f in features[100:]]
for tl in tlabels:
assert tl in labelset
def test_one_against_one():
M = milk.supervised.multi.one_against_one(base)
M = M.train(features[:100,:],labels[:100])
tlabels = [M.apply(f) for f in features[100:]]
for tl in tlabels:
assert tl in labelset
tlabels_many = M.apply_many(features[100:])
ri = celltypes_rna.index(ct)
cur.append(R[:,ri])
except ValueError:
pass
if not len(cur):
raise KeyError("?")
features.append(np.array(cur).mean(0))
selected = np.array(selected)
features = np.array(features)
gosweights = generate_gosweights(features, gene2ensembl, 'maxabs', ['molecular_function', 'biological_process'])
gosweights = thresh_features(gosweights)
learner = norm_learner(ctransforms(remove_constant_features(), zscore_normalise(), select_learner(12), lasso_relaxed(.000225010113525, .1)), 0)
rna_ge_gosweigths_mpbf_ma,labels = rna_ge_gosweigths.f('maxabs', ['molecular_function', 'biological_process'])
rna_ge_gosweigths_mpbf_ma = thresh_features(rna_ge_gosweigths_mpbf_ma)
model = learner.train(rna_ge_gosweigths_mpbf_ma, labels)
drugvalues, celltypes, drugs = read_training()
olines = read_olines()
results = []
for o in olines:
if o in testing:
oi = allct.index(o)
r = model.apply(gosweights[oi])
else:
oi = celltypes.index(o)
import numpy as np
import random
import milk.supervised.svm
import milk.supervised.multi
from milk.supervised.classifier import ctransforms
from fast_classifier import fast_classifier
import milksets.wine
features,labels = milksets.wine.load()
A = np.arange(len(features))
random.seed(9876543210)
random.shuffle(A)
features = features[A]
labels = labels[A]
labelset = set(labels)
base = ctransforms(milk.supervised.svm.svm_raw(C=2.,kernel=milk.supervised.svm.rbf_kernel(2.**-3)),milk.supervised.svm.svm_binary())
def test_one_against_rest():
M = milk.supervised.multi.one_against_rest(base)
M = M.train(features[:100,:],labels[:100])
tlabels = [M.apply(f) for f in features[100:]]
for tl in tlabels:
assert tl in labelset
def test_one_against_one():
M = milk.supervised.multi.one_against_one(base)
M = M.train(features[:100,:],labels[:100])
tlabels = [M.apply(f) for f in features[100:]]
for tl in tlabels:
assert tl in labelset
('prune(gow, .95)', prune_features(rna_ge_gosweigths_add, .95)),
('prune(gow-ma, .95)', prune_features(rna_ge_gosweigths_maxabs, .95)),
]:
features = data[0]
labels = data[1]
for name,learner in [
('random(1)', random_learner(1)),
('random(2)', random_learner(2)),
('random(3)', random_learner(3)),
('ridge(1.)', ridge_regression(1.)),
('ridge(.128)', ridge_regression(.128)),
('ridge(.001)', ridge_regression(.001)),
('ridge(1e-6)', ridge_regression(1e-6)),
('select+ridge(.001)', ctransforms(remove_constant_features(), select_learner(12), ridge_regression(.001))),
('relaxed', lasso_relaxed(.000225010113525, .1)),
('sel-relaxed', ctransforms(remove_constant_features(), select_learner(12), lasso_relaxed(.000225010113525, .1))),
('zs-sel-relaxed', ctransforms(remove_constant_features(), zscore_normalise(), select_learner(12), lasso_relaxed(.000225010113525, .1))),
('sel+lasso(1e-7)', ctransforms(remove_constant_features(), select_learner(12), lasso_regression(1e-7))),
('sel+lasso(1e-5)', ctransforms(remove_constant_features(), select_learner(16), lasso_regression(1e-5))),
('zs+lasso(1e-5)', ctransforms(remove_constant_features(), zscore_normalise(), lasso_regression_guess())),
('min0+lasso(1e-5)', ctransforms(remove_constant_features(), min_to_zero(), lasso_regression_guess())),
('sel+lasso(0.000225010113525)', ctransforms(remove_constant_features(), select_learner(16), lasso_regression(0.000225010113525))),
('zs+sel+lasso(0.000225010113525)', ctransforms(remove_constant_features(), zscore_normalise(), select_learner(16), lasso_regression(0.000225010113525))),
('zs+sel+lasso_learn', ctransforms(remove_constant_features(), zscore_normalise(), select_learner(12), lasso_regression_with_learning())),
('zs+sel+lasso_guess', ctransforms(remove_constant_features(), zscore_normalise(), select_learner(12), lasso_regression_guess())),
('zs+lasso_guess', ctransforms(remove_constant_features(), zscore_normalise(), lasso_regression_guess())),
('zs+sel+min0+lasso_guess', ctransforms(remove_constant_features(), zscore_normalise(), select_learner(12), min_to_zero(), lasso_regression_guess())),
('zs+min0+lasso_guess', ctransforms(remove_constant_features(), zscore_normalise(), min_to_zero(), lasso_regression_guess())),
