Skip to content

[MRG+1] Enforce deterministic output in kernel PCA #13241

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 8 commits into from
Feb 26, 2019
Merged

[MRG+1] Enforce deterministic output in kernel PCA #13241

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
3986762
enforce deterministic output in kernel PCA
bellet Feb 25, 2019
3b1590f
add tests and update whats new
bellet Feb 25, 2019
2ebeb98
replace state by rng
bellet Feb 25, 2019
348d9a7
simplified assert
bellet Feb 25, 2019
3370c21
avoid copy
bellet Feb 25, 2019
0e1c1df
clarify tests
bellet Feb 25, 2019
d4b46ee
remove now useless comment
bellet Feb 25, 2019
ac5f543
use rng as seed everywhere
bellet Feb 25, 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
7 changes: 4 additions & 3 deletions doc/whats_new/v0.21.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -76,16 +76,17 @@ Support for Python 3.4 and below has been officially dropped.
the default value is used.
:issue:`12988` by :user:`Zijie (ZJ) Poh <zjpoh>`.

:mod:`sklearn.decomposition`
............................

- |Fix| Fixed a bug in :class:`decomposition.NMF` where `init = 'nndsvd'`,
`init = 'nndsvda'`, and `init = 'nndsvdar'` are allowed when
`n_components < n_features` instead of
`n_components <= min(n_samples, n_features)`.
:issue:`11650` by :user:`Hossein Pourbozorg <hossein-pourbozorg>` and
:user:`Zijie (ZJ) Poh <zjpoh>`.

- |Enhancement| :class:`decomposition.KernelPCA` now has deterministic output
(resolved sign ambiguity in eigenvalue decomposition of the kernel matrix).
:issue:`13241` by :user:`Aurélien Bellet <bellet>`.

:mod:`sklearn.discriminant_analysis`
....................................

Expand Down
5 changes: 5 additions & 0 deletions sklearn/decomposition/kernel_pca.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,6 +8,7 @@
from scipy.sparse.linalg import eigsh

from ..utils import check_random_state
from ..utils.extmath import svd_flip
from ..utils.validation import check_is_fitted, check_array
from ..exceptions import NotFittedError
from ..base import BaseEstimator, TransformerMixin, _UnstableOn32BitMixin
Expand Down Expand Up @@ -210,6 +211,10 @@ def _fit_transform(self, K):
maxiter=self.max_iter,
v0=v0)

# flip eigenvectors' sign to enforce deterministic output
self.alphas_, _ = svd_flip(self.alphas_,
np.empty_like(self.alphas_).T)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

awesome, thanks! maybe remove the note: above since it's no longer relevant.


# sort eigenvectors in descending order
indices = self.lambdas_.argsort()[::-1]
self.lambdas_ = self.lambdas_[indices]
Expand Down
17 changes: 16 additions & 1 deletion sklearn/decomposition/tests/test_kernel_pca.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,7 @@

from sklearn.utils.testing import (assert_array_almost_equal, assert_less,
assert_equal, assert_not_equal,
assert_raises)
assert_raises, assert_allclose)

from sklearn.decomposition import PCA, KernelPCA
from sklearn.datasets import make_circles
Expand Down Expand Up @@ -71,6 +71,21 @@ def test_kernel_pca_consistent_transform():
assert_array_almost_equal(transformed1, transformed2)


def test_kernel_pca_deterministic_output():
rng = np.random.RandomState(0)
X = rng.rand(10, 10)
eigen_solver = ('arpack', 'dense')

for solver in eigen_solver:
transformed_X = np.zeros((20, 2))
for i in range(20):
kpca = KernelPCA(n_components=2, eigen_solver=solver,
random_state=rng)
transformed_X[i, :] = kpca.fit_transform(X)[0]
assert_allclose(
transformed_X, np.tile(transformed_X[0, :], 20).reshape(20, 2))


def test_kernel_pca_sparse():
rng = np.random.RandomState(0)
X_fit = sp.csr_matrix(rng.random_sample((5, 4)))
Expand Down
14 changes: 14 additions & 0 deletions sklearn/decomposition/tests/test_pca.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,7 @@

from sklearn.utils.testing import assert_almost_equal
from sklearn.utils.testing import assert_array_almost_equal
from sklearn.utils.testing import assert_allclose
from sklearn.utils.testing import assert_equal
from sklearn.utils.testing import assert_greater
from sklearn.utils.testing import assert_raise_message
Expand Down Expand Up @@ -703,6 +704,19 @@ def test_pca_dtype_preservation(svd_solver):
check_pca_int_dtype_upcast_to_double(svd_solver)


def test_pca_deterministic_output():
rng = np.random.RandomState(0)
X = rng.rand(10, 10)

for solver in solver_list:
transformed_X = np.zeros((20, 2))
for i in range(20):
pca = PCA(n_components=2, svd_solver=solver, random_state=rng)
transformed_X[i, :] = pca.fit_transform(X)[0]
assert_allclose(
transformed_X, np.tile(transformed_X[0, :], 20).reshape(20, 2))


def check_pca_float_dtype_preservation(svd_solver):
# Ensure that PCA does not upscale the dtype when input is float32
X_64 = np.random.RandomState(0).rand(1000, 4).astype(np.float64)
Expand Down