66from .metrics .scorer import _deprecate_loss_and_score_funcs
77
88def learning_curve (estimator , X , y ,
9- n_samples_range = np .arange (0.1 , 1.1 , 0.1 ), cv = None , scoring = None ,
9+ n_samples_range = np .linspace (0.1 , 1.0 , 10 ), cv = None , scoring = None ,
1010 n_jobs = 1 , verbose = False , random_state = None ):
1111 """ TODO document me
1212 Parameters
@@ -28,7 +28,7 @@ def learning_curve(estimator, X, y,
2828 n_samples_range = np .asarray (n_samples_range )
2929 n_min_required_samples = np .min (n_samples_range )
3030 n_max_required_samples = np .max (n_samples_range )
31- if np .issubdtype (n_samples_range .dtype , float ):
31+ if np .issubdtype (n_samples_range .dtype , np . float ):
3232 if n_min_required_samples <= 0.0 or n_max_required_samples > 1.0 :
3333 raise ValueError ("n_samples_range must be within ]0, 1], "
3434 "but is within [%f, %f]."
@@ -61,19 +61,22 @@ def learning_curve(estimator, X, y,
6161 "does not." % estimator )
6262 scorer = _deprecate_loss_and_score_funcs (scoring = scoring )
6363
64- scores = []
65- for n_train_samples in n_samples_range :
66- out = Parallel (
67- # TODO set pre_dispatch parameter? what is it good for?
68- n_jobs = n_jobs , verbose = verbose )(
69- delayed (_fit_estimator )(
70- estimator , X , y , train [:n_train_samples ], test , scorer ,
71- verbose )
72- for train , test in cv )
73- scores .append (np .mean (out , axis = 0 ))
74- scores = np .array (scores )
64+ out = Parallel (
65+ # TODO use pre_dispatch parameter? what is it good for?
66+ n_jobs = n_jobs , verbose = verbose )(
67+ delayed (_fit_estimator )(
68+ estimator , X , y , train [:n_train_samples ], test , scorer ,
69+ verbose )
70+ for train , test in cv for n_train_samples in n_samples_range )
7571
76- return n_samples_range , scores [:, 0 ], scores [:, 1 ]
72+ out = np .asarray (out )
73+ train_scores = np .zeros (n_samples_range .shape , dtype = np .float )
74+ test_scores = np .zeros (n_samples_range .shape , dtype = np .float )
75+ for i , n_train_samples in enumerate (n_samples_range ):
76+ res_indices = np .where (out [:, 0 ] == n_train_samples )
77+ train_scores [i ], test_scores [i ] = out [res_indices [0 ], 1 :].mean (axis = 0 )
78+
79+ return n_samples_range , train_scores , test_scores
7780
7881def _fit_estimator (base_estimator , X , y , train , test , scorer , verbose ):
7982 # TODO similar to fit_grid_point from grid search, refactor
@@ -85,4 +88,4 @@ def _fit_estimator(base_estimator, X, y, train, test, scorer, verbose):
8588 else :
8689 train_score = scorer (estimator , X [train ], y [train ])
8790 test_score = scorer (estimator , X [test ], y [test ])
88- return train_score , test_score
91+ return train . shape [ 0 ], train_score , test_score
0 commit comments