8000 [MRG] Save sample_weight_arr instead of sample_weight in KernelDensity by aditya1702 · Pull Request #13772 · scikit-learn/scikit-learn · GitHub
[go: up one dir, main page]
Skip to content

[MRG] Save sample_weight_arr instead of sample_weight in KernelDensity #13772

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 15 commits into from
May 14, 2019
Merged

[MRG] Save sample_weight_arr instead of sample_weight in KernelDensity #13772

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
15 commits
Select commit Hold shift + click to select a range
c199cb6
change sample_weight to sample_weight_arr
aditya1702 May 2, 2019
3933e99
add a test
aditya1702 May 3, 2019
0476cf3
Fix 84 characters in a line
aditya1702 May 3, 2019
c136f2a
add one more line between functions
aditya1702 May 3, 2019
b6c483a
set sample_state_arr
aditya1702 May 3, 2019
4baab0d
assign sample weights after unpickling
aditya1702 May 3, 2019
7e51d13
save sample_weight and sum_weight too
aditya1702 May 3, 2019
2228155
set sample_weight and sum_weight after unpickling
aditya1702 May 3, 2019
1ce2496
also handle the case when sample_weight is None
aditya1702 May 3, 2019
9442dab
initialise n_samples
aditya1702 May 3, 2019
2128a24
check if sample weight has proper size
aditya1702 May 3, 2019
d2e4581
Restore backward compat with 0.20 pickle files + factor redundant code
ogrisel May 13, 2019
57c3700
Merge branch 'master' into HEAD
jnothman May 13, 2019
3995316
DOC add what's new
jnothman May 13, 2019
5539ff8
Factorize KDE pickling tests via pytest.mark.parametrize
ogrisel May 14, 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: 8 additions & 1 deletion doc/whats_new/v0.21.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,7 @@
Version 0.21.1
==============

**May 2019**
**17 May 2019**


This is a bug-fix release with some minor documentation improvements and
Expand All @@ -24,6 +24,13 @@ Changelog
``Y == None``.
:issue:`13864` by :user:`Paresh Mathur <rick2047>`.

:mod:`sklearn.neighbors`
......................

- |Fix| Fixed a bug in :class:`neighbors.KernelDensity` which could not be
restored from a pickle if ``sample_weight`` had been used.
:issue:`13772` by :user:`Aditya Vyas <aditya1702>`.


.. _changes_0_21:

Expand Down
77 changes: 44 additions & 33 deletions sklearn/neighbors/binary_tree.pxi
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1064,10 +1064,17 @@ cdef class BinaryTree:

def __init__(self, data,
leaf_size=40, metric='minkowski', sample_weight=None, **kwargs):
self.data_arr = np.asarray(data, dtype=DTYPE, order='C')
self.data = get_memview_DTYPE_2D(self.data_arr)
# validate data
if data.size == 0:
raise ValueError("X is an empty array")

if leaf_size < 1:
raise ValueError("leaf_size must be greater than or equal to 1")

n_samples = data.shape[0]
n_features = data.shape[1]

self.data_arr = np.asarray(data, dtype=DTYPE, order='C')
self.leaf_size = leaf_size
self.dist_metric = DistanceMetric.get_metric(metric, **kwargs)
self.euclidean = (self.dist_metric.__class__.__name__
Expand All @@ -1079,26 +1086,6 @@ cdef class BinaryTree:
'{BinaryTree}'.format(metric=metric,
**DOC_DICT))

# validate data
if self.data.size == 0:
raise ValueError("X is an empty array")

if leaf_size < 1:
raise ValueError("leaf_size must be greater than or equal to 1")

n_samples = self.data.shape[0]
n_features = self.data.shape[1]


if sample_weight is not None:
self.sample_weight_arr = np.asarray(sample_weight, dtype=DTYPE, order='C')
self.sample_weight = get_memview_DTYPE_1D(self.sample_weight_arr)
self.sum_weight = np.sum(self.sample_weight)
else:
self.sample_weight = None
self.sum_weight = <DTYPE_t> n_samples


# determine number of levels in the tree, and from this
# the number of nodes in the tree. This results in leaf nodes
# with numbers of points between leaf_size and 2 * leaf_size
Expand All @@ -1107,15 +1094,34 @@ cdef class BinaryTree:

# allocate arrays for storage
self.idx_array_arr = np.arange(n_samples, dtype=ITYPE)
self.idx_array = get_memview_ITYPE_1D(self.idx_array_arr)

self.node_data_arr = np.zeros(self.n_nodes, dtype=NodeData)
self.node_data = get_memview_NodeData_1D(self.node_data_arr)

self._update_sample_weight(n_samples, sample_weight)
self._update_memviews()

# Allocate tree-specific data
allocate_data(self, self.n_nodes, n_features)
self._recursive_build(0, 0, n_samples)

def _update_sample_weight(self, n_samples, sample_weight):
if sample_weight is not None:
self.sample_weight_arr = np.asarray(
sample_weight, dtype=DTYPE, order='C')
self.sample_weight = get_memview_DTYPE_1D(
self.sample_weight_arr)
self.sum_weight = np.sum(self.sample_weight)
else:
self.sample_weight = None
self.sample_weight_arr = np.empty(1, dtype=DTYPE, order='C')
self.sum_weight = <DTYPE_t> n_samples

def _update_memviews(self):
self.data = get_memview_DTYPE_2D(self.data_arr)
self.idx_array = get_memview_ITYPE_1D(self.idx_array_arr)
self.node_data = get_memview_NodeData_1D(self.node_data_arr)
self.node_bounds = get_memview_DTYPE_3D(self.node_bounds_arr)


def __reduce__(self):
"""
reduce method used for pickling
Expand All @@ -1126,6 +1132,13 @@ cdef class BinaryTree:
"""
get state for pickling
"""
if self.sample_weight is not None:
# pass the numpy array
sample_weight_arr = self.sample_weight_arr
else:
# pass None to avoid confusion with the empty place holder
# of size 1 from __cinit__
sample_weight_arr = None
return (self.data_arr,
self.idx_array_arr,
self.node_data_arr,
Expand All @@ -1138,7 +1151,7 @@ cdef class BinaryTree:
int(self.n_splits),
int(self.n_calls),
self.dist_metric,
self.sample_weight)
sample_weight_arr)

def __setstate__(self, state):
"""
Expand All @@ -1148,12 +1161,6 @@ cdef class BinaryTree:
self.idx_array_arr = state[1]
self.node_data_arr = state[2]
self.node_bounds_arr = state[3]

self.data = get_memview_DTYPE_2D(self.data_arr)
self.idx_array = get_memview_ITYPE_1D(self.idx_array_arr)
self.node_data = get_memview_NodeData_1D(self.node_data_arr)
self.node_bounds = get_memview_DTYPE_3D(self.node_bounds_arr)

self.leaf_size = state[4]
self.n_levels = state[5]
self.n_nodes = state[6]
Expand All @@ -1162,9 +1169,13 @@ cdef class BinaryTree:
self.n_splits = state[9]
self.n_calls = state[10]
self.dist_metric = state[11]
sample_weight_arr = state[12]

self.euclidean = (self.dist_metric.__class__.__name__
== 'EuclideanDistance')
self.sample_weight = state[12]
n_samples = self.data_arr.shape[0]
self._update_sample_weight(n_samples, sample_weight_arr)
self._update_memviews()

def get_tree_stats(self):
return (self.n_trims, self.n_leaves, self.n_splits)
Expand Down
5 changes: 3 additions & 2 deletions sklearn/neighbors/tests/test_kde.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -205,14 +205,15 @@ def test_kde_sample_weights():
assert_allclose(scores_scaled_weight, scores_weight)


def test_pickling(tmpdir):
@pytest.mark.parametrize('sample_weight', [None, [0.1, 0.2, 0.3]])
def test_pickling(tmpdir, sample_weight):
# Make sure that predictions are the same before and after pickling. Used
# to be a bug because sample_weights wasn't pickled and the resulting tree
# would miss some info.

kde = KernelDensity()
data = np.reshape([1., 2., 3.], (-1, 1))
kde.fit(data)
kde.fit(data, sample_weight=sample_weight)

X = np.reshape([1.1, 2.1], (-1, 1))
scores = kde.score_samples(X)
Expand Down
0