def test_precluster():
learner = precluster_learner([2], base=fast_classifier(), R=12)
features, labels = gen_data(22)
model = learner.train(features, labels)
assert model.apply([c0() for i in xrange(35)])
assert not model.apply([c1() for i in xrange(35)])
def test_precluster():
learner = precluster_learner([2], base=fast_classifier(), R=12)
features, labels = gen_data(22)
model = learner.train(features,labels)
assert model.apply([c0() for i in xrange(35)])
assert not model.apply([c1() for i in xrange(35)])
def test_gridminimise_return():
from milksets.wine import load
features,labels = load()
learner = fast_classifier()
gridminimise(learner, features, labels, { 'ignore' : [0] })
_,error = gridminimise(learner, features, labels, { 'ignore' : [0] }, return_value=True, nfolds=5)
cmat,_ = milk.nfoldcrossvalidation(features, labels, learner=learner, nfolds=5)
assert error == cmat.sum()-cmat.trace()
def test_codebook_learner_case1():
learner = select_precluster([2], base=fast_classifier())
learner.rmax = 1
features, labels = gen_data(23, 1)
model = learner.train(features,labels)
assert model.apply(([c0() for i in xrange(35)],[]))
assert not model.apply(([c1() for i in xrange(35)],[]))
def test_codebook_learner_case1():
learner = select_precluster([2], base=fast_classifier())
learner.rmax = 1
features, labels = gen_data(23, 1)
model = learner.train(features, labels)
assert model.apply(([c0() for i in xrange(35)], []))
assert not model.apply(([c1() for i in xrange(35)], []))
def test_gridminimise_return():
from milksets.wine import load
features, labels = load()
learner = fast_classifier()
gridminimise(learner, features, labels, {'ignore': [0]})
_, error = gridminimise(learner,
features,
labels, {'ignore': [0]},
return_value=True,
nfolds=5)
cmat, _ = milk.nfoldcrossvalidation(features,
labels,
learner=learner,
nfolds=5)
assert error == cmat.sum() - cmat.trace()
def test_one_by_one():
np.random.seed(23)
r = np.random.random
ps = np.array([.7,.5,.8,.3,.8])
learner = milk.supervised.multi_label.one_by_one(fast_classifier())
universe = list(range(len(ps)))
for _ in range(10):
labels = []
features = []
bases = [np.random.rand(20) for pj in ps]
for i in range(256):
cur = []
curf = np.zeros(20,float)
for j,pj in enumerate(ps):
if r() < pj:
cur.append(j)
curf += r()*bases[j]
if not cur: continue
labels.append(cur)
features.append(curf)
model = learner.train(features, labels)
predicted = model.apply_many(features)
matrix = np.zeros((2,2), int)
for t,p in zip(labels, predicted):
for ell in universe:
row = (ell in t)
col = (ell in p)
matrix[row,col] += 1
Tn,Fp = matrix[0]
Fn,Tp = matrix[1]
prec = Tp/float(Tp+Fp)
recall = Tp/float(Tp+Fn)
F1 = 2*prec*recall/(prec + recall)
assert F1 > .3
def test_one_by_one():
np.random.seed(23)
r = np.random.random
ps = np.array([.7, .5, .8, .3, .8])
learner = milk.supervised.multi_label.one_by_one(fast_classifier())
universe = range(len(ps))
for _ in xrange(10):
labels = []
features = []
bases = [np.random.rand(20) for pj in ps]
for i in xrange(256):
cur = []
curf = np.zeros(20, float)
for j, pj in enumerate(ps):
if r() < pj:
cur.append(j)
curf += r() * bases[j]
if not cur: continue
labels.append(cur)
features.append(curf)
model = learner.train(features, labels)
predicted = model.apply_many(features)
matrix = np.zeros((2, 2), int)
for t, p in zip(labels, predicted):
for ell in universe:
row = (ell in t)
col = (ell in p)
matrix[row, col] += 1
Tn, Fp = matrix[0]
Fn, Tp = matrix[1]
prec = Tp / float(Tp + Fp)
recall = Tp / float(Tp + Fn)
F1 = 2 * prec * recall / (prec + recall)
assert F1 > .3
def test_precluster():
np.random.seed(22)
learner = precluster_learner([2], base=fast_classifier(), R=12)
def c0():
return np.random.rand(8)
def c1():
return c0()+2.*np.ones(8)
features = []
labels =[]
for i in xrange(200):
f = []
for j in xrange(40):
use_0 = (i < 100 and j < 30) or (i >= 100 and j >= 30)
if use_0: f.append(c0())
else: f.append(c1())
labels.append((i < 100))
features.append(f)
model = learner.train(features,labels)
assert model.apply([c0() for i in xrange(35)])
assert not model.apply([c1() for i in xrange(35)])
def test_precluster():
np.random.seed(22)
learner = precluster_learner([2], base=fast_classifier(), R=12)
def c0():
return np.random.rand(8)
def c1():
return c0() + 2. * np.ones(8)
features = []
labels = []
for i in xrange(200):
f = []
for j in xrange(40):
use_0 = (i < 100 and j < 30) or (i >= 100 and j >= 30)
if use_0: f.append(c0())
else: f.append(c1())
labels.append((i < 100))
features.append(f)
model = learner.train(features, labels)
assert model.apply([c0() for i in xrange(35)])
assert not model.apply([c1() for i in xrange(35)])
import milk.ext.jugparallel
from milksets.wine import load
from milk.tests.fast_classifier import fast_classifier
features, labels = load()
classified = milk.ext.jugparallel.nfoldcrossvalidation(
features, labels, learner=fast_classifier())
classified_wpred = milk.ext.jugparallel.nfoldcrossvalidation(
features, labels, learner=fast_classifier(), return_predictions=True)
import milk.ext.jugparallel from milksets.wine import load from milk.tests.fast_classifier import fast_classifier features,labels = load() classified = milk.ext.jugparallel.nfoldcrossvalidation(features, labels, learner=fast_classifier()) classified_wpred = milk.ext.jugparallel.nfoldcrossvalidation(features, labels, learner=fast_classifier(), return_predictions=True)
