Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Commit 325130f

Browse files
jjerphanogriseljeremiedbbthomasjpfan
authored
MAINT Create private _pairwise_distances_reductions submodule (#23724)
Co-authored-by: Olivier Grisel <[email protected]> Co-authored-by: Jérémie du Boisberranger <[email protected]> Co-authored-by: Thomas J. Fan <[email protected]> Co-authored-by: Jérémie du Boisberranger <[email protected]>
1 parent 9dbbdbd commit 325130f

18 files changed

+2593
-2186
lines changed

‎sklearn/metrics/_dist_metrics.pxd.tp

Lines changed: 0 additions & 20 deletions
Original file line numberDiff line numberDiff line change
@@ -101,23 +101,3 @@ cdef class DistanceMetric{{name_suffix}}:
101101
cdef DTYPE_t _dist_to_rdist(self, {{DTYPE_t}} dist) nogil except -1
102102

103103
{{endfor}}
104-
105-
######################################################################
106-
# DatasetsPair base class
107-
cdef class DatasetsPair:
108-
cdef DistanceMetric distance_metric
109-
110-
cdef ITYPE_t n_samples_X(self) nogil
111-
112-
cdef ITYPE_t n_samples_Y(self) nogil
113-
114-
cdef DTYPE_t dist(self, ITYPE_t i, ITYPE_t j) nogil
115-
116-
cdef DTYPE_t surrogate_dist(self, ITYPE_t i, ITYPE_t j) nogil
117-
118-
119-
cdef class DenseDenseDatasetsPair(DatasetsPair):
120-
cdef:
121-
const DTYPE_t[:, ::1] X
122-
const DTYPE_t[:, ::1] Y
123-
ITYPE_t d

‎sklearn/metrics/_dist_metrics.pyx.tp

Lines changed: 0 additions & 161 deletions
Original file line numberDiff line numberDiff line change
@@ -32,7 +32,6 @@ implementation_specific_values = [
3232

3333
import numpy as np
3434
cimport numpy as cnp
35-
from cython cimport final
3635

3736
cnp.import_array() # required in order to use C-API
3837

@@ -1171,163 +1170,3 @@ cdef class PyFuncDistance{{name_suffix}}(DistanceMetric{{name_suffix}}):
11711170
"vectors and return a float.")
11721171

11731172
{{endfor}}
1174-
1175-
######################################################################
1176-
# Datasets Pair Classes
1177-
cdef class DatasetsPair:
1178-
"""Abstract class which wraps a pair of datasets (X, Y).
1179-
1180-
This class allows computing distances between a single pair of rows of
1181-
of X and Y at a time given the pair of their indices (i, j). This class is
1182-
specialized for each metric thanks to the :func:`get_for` factory classmethod.
1183-
1184-
The handling of parallelization over chunks to compute the distances
1185-
and aggregation for several rows at a time is done in dedicated
1186-
subclasses of PairwiseDistancesReduction that in-turn rely on
1187-
subclasses of DatasetsPair for each pair of rows in the data. The goal
1188-
is to make it possible to decouple the generic parallelization and
1189-
aggregation logic from metric-specific computation as much as
1190-
possible.
1191-
1192-
X and Y can be stored as C-contiguous np.ndarrays or CSR matrices
1193-
in subclasses.
1194-
1195-
This class avoids the overhead of dispatching distance computations
1196-
to :class:`sklearn.metrics.DistanceMetric` based on the physical
1197-
representation of the vectors (sparse vs. dense). It makes use of
1198-
cython.final to remove the overhead of dispatching method calls.
1199-
1200-
Parameters
1201-
----------
1202-
distance_metric: DistanceMetric
1203-
The distance metric responsible for computing distances
1204-
between two vectors of (X, Y).
1205-
"""
1206-
1207-
@classmethod
1208-
def get_for(
1209-
cls,
1210-
X,
1211-
Y,
1212-
str metric="euclidean",
1213-
dict metric_kwargs=None,
1214-
) -> DatasetsPair:
1215-
"""Return the DatasetsPair implementation for the given arguments.
1216-
1217-
Parameters
1218-
----------
1219-
X : {ndarray, sparse matrix} of shape (n_samples_X, n_features)
1220-
Input data.
1221-
If provided as a ndarray, it must be C-contiguous.
1222-
If provided as a sparse matrix, it must be in CSR format.
1223-
1224-
Y : {ndarray, sparse matrix} of shape (n_samples_Y, n_features)
1225-
Input data.
1226-
If provided as a ndarray, it must be C-contiguous.
1227-
If provided as a sparse matrix, it must be in CSR format.
1228-
1229-
metric : str, default='euclidean'
1230-
The distance metric to compute between rows of X and Y.
1231-
The default metric is a fast implementation of the Euclidean
1232-
metric. For a list of available metrics, see the documentation
1233-
of :class:`~sklearn.metrics.DistanceMetric`.
1234-
1235-
metric_kwargs : dict, default=None
1236-
Keyword arguments to pass to specified metric function.
1237-
1238-
Returns
1239-
-------
1240-
datasets_pair: DatasetsPair
1241-
The suited DatasetsPair implementation.
1242-
"""
1243-
cdef:
1244-
DistanceMetric distance_metric = DistanceMetric.get_metric(
1245-
metric,
1246-
**(metric_kwargs or {})
1247-
)
1248-
1249-
if not(X.dtype == Y.dtype == np.float64):
1250-
raise ValueError(
1251-
f"Only 64bit float datasets are supported at this time, "
1252-
f"got: X.dtype={X.dtype} and Y.dtype={Y.dtype}."
1253-
)
1254-
1255-
# Metric-specific checks that do not replace nor duplicate `check_array`.
1256-
distance_metric._validate_data(X)
1257-
distance_metric._validate_data(Y)
1258-
1259-
# TODO: dispatch to other dataset pairs for sparse support once available:
1260-
if issparse(X) or issparse(Y):
1261-
raise ValueError("Only dense datasets are supported for X and Y.")
1262-
1263-
return DenseDenseDatasetsPair(X, Y, distance_metric)
1264-
1265-
def __init__(self, DistanceMetric distance_metric):
1266-
self.distance_metric = distance_metric
1267-
1268-
cdef ITYPE_t n_samples_X(self) nogil:
1269-
"""Number of samples in X."""
1270-
# This is a abstract method.
1271-
# This _must_ always be overwritten in subclasses.
1272-
# TODO: add "with gil: raise" here when supporting Cython 3.0
1273-
return -999
1274-
1275-
cdef ITYPE_t n_samples_Y(self) nogil:
1276-
"""Number of samples in Y."""
1277-
# This is a abstract method.
1278-
# This _must_ always be overwritten in subclasses.
1279-
# TODO: add "with gil: raise" here when supporting Cython 3.0
1280-
return -999
1281-
1282-
cdef DTYPE_t surrogate_dist(self, ITYPE_t i, ITYPE_t j) nogil:
1283-
return self.dist(i, j)
1284-
1285-
cdef DTYPE_t dist(self, ITYPE_t i, ITYPE_t j) nogil:
1286-
# This is a abstract method.
1287-
# This _must_ always be overwritten in subclasses.
1288-
# TODO: add "with gil: raise" here when supporting Cython 3.0
1289-
return -1
1290-
1291-
@final
1292-
cdef class DenseDenseDatasetsPair(DatasetsPair):
1293-
"""Compute distances between row vectors of two arrays.
1294-
1295-
Parameters
1296-
----------
1297-
X: ndarray of shape (n_samples_X, n_features)
1298-
Rows represent vectors. Must be C-contiguous.
1299-
1300-
Y: ndarray of shape (n_samples_Y, n_features)
1301-
Rows represent vectors. Must be C-contiguous.
1302-
1303-
distance_metric: DistanceMetric
1304-
The distance metric responsible for computing distances
1305-
between two row vectors of (X, Y).
1306-
"""
1307-
1308-
def __init__(self, X, Y, DistanceMetric distance_metric):
1309-
super().__init__(distance_metric)
1310-
# Arrays have already been checked
1311-
self.X = X
1312-
self.Y = Y
1313-
self.d = X.shape[1]
1314-
1315-
@final
1316-
cdef ITYPE_t n_samples_X(self) nogil:
1317-
return self.X.shape[0]
1318-
1319-
@final
1320-
cdef ITYPE_t n_samples_Y(self) nogil:
1321-
return self.Y.shape[0]
1322-
1323-
@final
1324-
cdef DTYPE_t surrogate_dist(self, ITYPE_t i, ITYPE_t j) nogil:
1325-
return self.distance_metric.rdist(&self.X[i, 0],
1326-
&self.Y[j, 0],
1327-
self.d)
1328-
1329-
@final
1330-
cdef DTYPE_t dist(self, ITYPE_t i, ITYPE_t j) nogil:
1331-
return self.distance_metric.dist(&self.X[i, 0],
1332-
&self.Y[j, 0],
1333-
self.d)

0 commit comments

Comments
 (0)