def logistic_regression_large(self, alpha, mu):
X, w0, gt = self._problem(n=1000,
nlf=100,
lf_prior=0.2,
lf_mean=0.7,
lf_sd=0.25,
nf=5000,
f_prior=0.05,
f_mean=0.6,
f_sd=0.25)
l = LogReg()
l.train(X=X,
n_iter=1000,
tol=1e-4,
w0=w0,
sample=False,
alpha=alpha,
mu=mu,
rate=0.01,
verbose=True)
self.assertGreater(np.mean(np.sign(X.dot(l.w)) == gt), 0.85)
def test_logistic_regression_sample(self):
print("Running logistic regression test with sparse operations")
X, w0, gt = self._problem(n=500,
nlf=75,
lf_prior=0.3,
lf_mean=0.7,
lf_sd=0.25,
nf=1000,
f_prior=0.1,
f_mean=0.6,
f_sd=0.25)
mu = 1e-4
l_d = LogReg()
l_d.train(X=X,
n_iter=2500,
tol=1e-4,
w0=w0,
sample=False,
alpha=self.ridge,
mu=mu,
rate=0.01,
verbose=True)
l_s = LogReg()
l_s.train(X=X,
n_iter=2500,
tol=1e-4,
w0=w0,
sample=True,
n_samples=200,
alpha=self.ridge,
mu=mu,
rate=0.01,
verbose=True)
# Check sample marginals are close to deterministic solutio
ld, ls = odds_to_prob(X.dot(l_d.w)), odds_to_prob(X.dot(l_s.w))
self.assertLessEqual(
np.linalg.norm(ld - ls) / np.linalg.norm(ld), 0.05)
gen_model.train(L_train, n_iter=1000, rate=1e-5)
from snorkel.learning import NaiveBayes
gen_model = NaiveBayes()
gen_model.train(L_train, n_iter=1000, rate=1e-5)
train_marginals = gen_model.marginals(L_train)
from snorkel.learning import LogReg
from snorkel.learning_utils import RandomSearch, ListParameter, RangeParameter
iter_param = ListParameter('n_iter', [250, 500, 1000, 2000])
rate_param = RangeParameter('rate', 1e-4, 1e-2, step=0.75, log_base=10)
reg_param = RangeParameter('mu', 1e-8, 1e-2, step=1, log_base=10)
disc_model = LogReg()
%load_ext autoreload
%autoreload 2
%matplotlib inline
from snorkel import SnorkelSession
session = SnorkelSession()
from snorkel.models import CandidateSet
from snorkel.models import candidate_subclass
from snorkel.annotations import FeatureManager
feature_manager = FeatureManager()
entity = candidate_subclass('entity', ['entity1', 'entity2'])
dev = session.query(CandidateSet).filter(CandidateSet.name == 'Protein1 Development Candidates').one()
%time F_dev = feature_manager.update(session, dev, 'Train1 Features', False)
#%time F_train = feature_manager.load(session, train, 'Train Features')
from snorkel.annotations import LabelManager
label_manager = LabelManager()
%time L_train = label_manager.create(session, c, 'LF Labels', f=LFs)
L_train
from snorkel.learning import NaiveBayes
gen_model = NaiveBayes()
gen_model.train(L_train, n_iter=1000, rate=1e-5)
gen_model.save(session, 'Generative Params')
train_marginals = gen_model.marginals(L_train)
gen_model.w
from snorkel.learning import LogReg
from snorkel.learning_utils import RandomSearch, ListParameter, RangeParameter
iter_param = ListParameter('n_iter', [250, 500, 1000, 2000])
rate_param = RangeParameter('rate', 1e-4, 1e-2, step=0.75, log_base=10)
reg_param = RangeParameter('mu', 1e-8, 1e-2, step=1, log_base=10)
disc_model = LogReg()
%time F_dev = feature_manager.update(session, dev, 'Train Features', False)
searcher = RandomSearch(disc_model, F_train, train_marginals, 10, iter_param, rate_param, reg_param)