test_sgl_logistic.py

import numpy
import subgradients, subgradients_semisparse, blockwise_descent, blockwise_descent_semisparse

__author__ = 'Romain Tavenard romain.tavenard[at]univ-rennes2.fr'

n = 1000
d = 20
groups = numpy.array([0] * int(d / 2) + [1] * (d - int(d / 2)))
alpha = 0.
lbda = .1

X = numpy.random.randn(n, d)
secret_beta = numpy.random.randn(d)
ind_sparse = numpy.zeros((d, ))
for i in range(d):
    if groups[i] == 0 or i % 2 == 0:
        secret_beta[i] = 0
    if i % 2 != 0:
        ind_sparse[i] = 1

y = numpy.ones((n, ))
y[numpy.exp(numpy.dot(X, secret_beta)) < 1.] = 0

#model = subgradients.SGL(groups=groups, alpha=0., lbda=0.1)
#model = subgradients_semisparse.SGL(groups=groups, alpha=0.1, lbda=0.1, ind_sparse=ind_sparse)
#model = blockwise_descent.SGL(groups=groups, alpha=0., lbda=0.1)
model = blockwise_descent_semisparse.SGL_LogisticRegression(groups=groups, alpha=alpha, lbda=lbda,
                                                            ind_sparse=ind_sparse, max_iter_outer=500)

model.fit(X, y)
beta_hat = model.coef_

print(numpy.linalg.norm(secret_beta - beta_hat))
for i, (betai_hat, betai) in enumerate(zip(beta_hat, secret_beta)):
    print("Component %02d: %.4f | %.4f" % (i, betai_hat, betai))
print("Correct classification rate: %.3f" % (numpy.sum(model.predict(X) == y) / n))