@@ -140,6 +140,12 @@ Changelog
140
140
`'auto' ` in 1.3. `None ` and `'warn' ` will be removed in 1.3. :pr: `20145 ` by
141
141
:user: `murata-yu `.
142
142
143
+ :mod: `sklearn.feature_selection `
144
+ ................................
145
+
146
+ - |Efficiency | Improve runtime performance of :func: `feature_selection.chi2 `
147
+ with boolean arrays. :pr: `22235 ` by `Thomas Fan `_.
148
+
143
149
:mod: `sklearn.datasets `
144
150
.......................
145
151
@@ -234,6 +240,12 @@ Changelog
234
240
:class: `ensemble.ExtraTreesClassifier `.
235
241
:pr: `20803 ` by :user: `Brian Sun <bsun94> `.
236
242
243
+
244
+ :mod: `sklearn.feature_selection `
245
+ ................................
246
+ - |Efficiency | Reduced memory usage of :func: `feature_selection.chi2 `.
247
+ :pr: `21837 ` by :user: `Louis Wagner <lrwagner> `
248
+
237
249
- |Efficiency | Fitting a :class: `ensemble.RandomForestClassifier `,
238
250
:class: `ensemble.RandomForestRegressor `, :class: `ensemble.ExtraTreesClassifier `,
239
251
:class: `ensemble.ExtraTreesRegressor `, and :class: `ensemble.RandomTreesEmbedding `
@@ -273,6 +285,13 @@ Changelog
273
285
F-statistic).
274
286
:pr: `17819 ` by :user: `Juan Carlos Alfaro Jiménez <alfaro96> `.
275
287
288
+ :mod: `sklearn.gaussian_process `
289
+ ...............................
290
+
291
+ - |Fix | :class: `gaussian_process.GaussianProcessClassifier ` raises
292
+ a more informative error if `CompoundKernel ` is passed via `kernel `.
293
+ :pr: `22223 ` by :user: `MarcoM <marcozzxx810> `.
294
+
276
295
:mod: `sklearn.impute `
277
296
.....................
278
297
@@ -322,6 +341,9 @@ Changelog
322
341
- |Enhancement | :class: `linear_model.QuantileRegressor ` support sparse input
323
342
for the highs based solvers.
324
343
:pr: `21086 ` by :user: `Venkatachalam Natchiappan <venkyyuvy> `.
344
+ In addition, those solvers now use the CSC matrix right from the
345
+ beginning which speeds up fitting.
346
+ :pr: `22206 ` by :user: `Christian Lorentzen <lorentzenchr> `.
325
347
326
348
- |Enhancement | Rename parameter `base_estimator ` to `estimator ` in
327
349
:class: `linear_model.RANSACRegressor ` to improve readability and consistency.
@@ -334,10 +356,20 @@ Changelog
334
356
:pr: `21481 ` by :user: `Guillaume Lemaitre <glemaitre> ` and
335
357
:user: `Andrés Babino <ababino> `.
336
358
359
+ - |Enhancement | :func: `linear_model.ElasticNet ` and
360
+ and other linear model classes using coordinate descent show error
361
+ messages when non-finite parameter weights are produced. :pr: `22148 `
362
+ by :user: `Christian Ritter <chritter> ` and :user: `Norbert Preining <norbusan> `.
363
+
337
364
- |Fix | :class: `linear_model.ElasticNetCV ` now produces correct
338
365
warning when `l1_ratio=0 `.
339
366
:pr: `21724 ` by :user: `Yar Khine Phyo <yarkhinephyo> `.
340
367
368
+ - |Enhancement | :class: `linear_model.ElasticNet ` and :class: `linear_model.Lasso `
369
+ now raise consistent error messages when passed invalid values for `l1_ratio `,
370
+ `alpha `, `max_iter ` and `tol `.
371
+ :pr: `22240 ` by :user: `Arturo Amor <ArturoAmorQ> `.
372
+
341
373
:mod: `sklearn.metrics `
342
374
......................
343
375
@@ -359,6 +391,11 @@ Changelog
359
391
A deprecation cycle was introduced.
360
392
:pr: `21576 ` by :user: `Paul-Emile Dugnat <pedugnat> `.
361
393
394
+ - |API | The `"wminkowski" ` metric of :class: `sklearn.metrics.DistanceMetric ` is deprecated
395
+ and will be removed in version 1.3. Instead the existing `"minkowski" ` metric now takes
396
+ in an optional `w ` parameter for weights. This deprecation aims at remaining consistent
397
+ with SciPy 1.8 convention. :pr: `21873 ` by :user: `Yar Khine Phyo <yarkhinephyo> `
398
+
362
399
- |Fix | :func: `metrics.silhouette_score ` now supports integer input for precomputed
363
400
distances. :pr: `22108 ` by `Thomas Fan `_.
364
401
@@ -382,6 +419,11 @@ Changelog
382
419
splits failed. Similarly raise an error during grid-search when the fits for
383
420
all the models and all the splits failed. :pr: `21026 ` by :user: `Loïc Estève <lesteve> `.
384
421
422
+ - |Enhancement | it is now possible to pass `scoring="matthews_corrcoef" ` to all
423
+ model selection tools with a `scoring ` argument to use the Matthews
424
+ correlation coefficient (MCC). :pr: `22203 ` by :user: `Olivier Grisel
425
+ <ogrisel> `.
426
+
385
427
- |Fix | :class: `model_selection.GridSearchCV `,
386
428
:class: `model_selection.HalvingGridSearchCV `
387
429
now validate input parameters in `fit ` instead of `__init__ `.
@@ -408,10 +450,23 @@ Changelog
408
450
ndarray with `np.nan ` when passed a `Float32 ` or `Float64 ` pandas extension
409
451
array with `pd.NA `. :pr: `21278 ` by `Thomas Fan `_.
410
452
453
+ - |Enhancement | Adds :term: `get_feature_names_out ` to
454
+ :class: `neighbors.RadiusNeighborsTransformer `, :class: `neighbors.KNeighborsTransformer `
455
+ and :class: `neighbors.NeighborhoodComponentsAnalysis `. :pr: `22212 ` by
456
+ :user : `Meekail Zain <micky774> `.
457
+
411
458
- |Fix | :class: `neighbors.KernelDensity ` now validates input parameters in `fit `
412
459
instead of `__init__ `. :pr: `21430 ` by :user: `Desislava Vasileva <DessyVV> ` and
413
460
:user: `Lucy Jimenez <LucyJimenez> `.
414
461
462
+ :mod: `sklearn.neural_network `
463
+ .............................
464
+
465
+ - |Enhancement | :func: `neural_network.MLPClassifier ` and
466
+ :func: `neural_network.MLPRegressor ` show error
467
+ messages when optimizers produce non-finite parameter weights. :pr: `22150 `
468
+ by :user: `Christian Ritter <chritter> ` and :user: `Norbert Preining <norbusan> `.
469
+
415
470
:mod: `sklearn.pipeline `
416
471
.......................
417
472
@@ -470,6 +525,12 @@ Changelog
470
525
parameters in `fit ` instead of `__init__ `.
471
526
:pr: `21436 ` by :user: `Haidar Almubarak <Haidar13 > `.
472
527
528
+ - |Enhancement | :func: `svm.SVR `, :func: `svm.SVC `, :func: `svm.NuSVR `,
529
+ :func: `svm.OneClassSVM `, :func: `svm.NuSVC ` now raise an error
530
+ when the dual-gap estimation produce non-finite parameter weights.
531
+ :pr: `22149 ` by :user: `Christian Ritter <chritter> ` and
532
+ :user: `Norbert Preining <norbusan> `.
533
+
473
534
:mod: `sklearn.utils `
474
535
....................
475
536
0 commit comments