Skip to content

[MRG] Use pseudo-inverse in Covariance #206

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 4 commits into from
Jun 5, 2019
Merged

[MRG] Use pseudo-inverse in Covariance #206

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
e2f319a
FIX: fix covariance algo
May 21, 2019
f57187f
some fixes and add non regression test
May 21, 2019
6cb9fc3
Use size instead of len
May 21, 2019
a8967ba
Address https://github.com/metric-learn/metric-learn/pull/206#pullreq…
Jun 5, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 5 additions & 4 deletions metric_learn/covariance.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,7 @@

from __future__ import absolute_import
import numpy as np
import scipy
from sklearn.base import TransformerMixin

from .base_metric import MahalanobisMixin
Expand All @@ -35,11 +36,11 @@ def fit(self, X, y=None):
y : unused
"""
X = self._prepare_inputs(X, ensure_min_samples=2)
M = np.cov(X, rowvar = False)
if M.ndim == 0:
M = 1./M
M = np.atleast_2d(np.cov(X, rowvar=False))
if M.size == 1:
M = 1. / M
else:
M = np.linalg.inv(M)
M = scipy.linalg.pinvh(M)

self.transformer_ = transformer_from_metric(np.atleast_2d(M))
return self
20 changes: 19 additions & 1 deletion test/metric_learn_test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,8 @@
from six.moves import xrange
from sklearn.metrics import pairwise_distances
from sklearn.datasets import load_iris, make_classification, make_regression
from numpy.testing import assert_array_almost_equal, assert_array_equal
from numpy.testing import (assert_array_almost_equal, assert_array_equal,
assert_allclose)
from sklearn.utils.testing import assert_warns_message
from sklearn.exceptions import ConvergenceWarning
from sklearn.utils.validation import check_X_y
Expand Down Expand Up @@ -53,6 +54,23 @@ def test_iris(self):
# deterministic result
self.assertAlmostEqual(csep, 0.72981476)

def test_singular_returns_pseudo_inverse(self):
"""Checks that if the input covariance matrix is singular, we return
the pseudo inverse"""
X, y = load_iris(return_X_y=True)
# We add a virtual column that is a linear combination of the other
# columns so that the covariance matrix will be singular
X = np.concatenate([X, X[:, :2].dot([[2], [3]])], axis=1)
cov_matrix = np.cov(X, rowvar=False)
covariance = Covariance()
covariance.fit(X)
pseudo_inverse = covariance.get_mahalanobis_matrix()
# here is the definition of a pseudo inverse according to wikipedia:
assert_allclose(cov_matrix.dot(pseudo_inverse).dot(cov_matrix),
cov_matrix)
assert_allclose(pseudo_inverse.dot(cov_matrix).dot(pseudo_inverse),
pseudo_inverse)


class TestLSML(MetricTestCase):
def test_iris(self):
Expand Down