U 2d<,@sddlZddlmZddlmZddlmZddlmZm Z ddl m Z ddl m Z ej dd Zejd d d ied fdd iedfgddZddZddZejdddgddZejd dddZddZd d!Zd"d#ZdS)$N) load_iris)DecisionTreeClassifier)shuffle)assert_allcloseassert_array_equal)learning_curve)LearningCurveDisplaycCsttddddiS)NT)Z return_X_y random_stater)rrr r K/tmp/pip-unpacked-wheel-zrfo1fqw/sklearn/model_selection/tests/test_plot.pydata sr zparams, err_type, err_msgstd_display_styleinvalidzUnknown std_display_style: score_typezUnknown score_type:c CsT|\}}tdd}dddg}tj||d tj|||fd|i|W5QRXdS) zCCheck that we raise a proper error when passing invalid parameters.rr 333333?333333??)match train_sizesN)rpytestZraisesrfrom_estimator) pyplotr paramsZerr_typeerr_msgXy estimatorrr r r 1test_learning_curve_display_parameters_validations   rcCs<|\}}tdd}dddg}tj||||d}ddl}|jdksDtt|jtsTt|jD]}t||j j sZtqZt|j tst|j D]&} t| |j j st| dkstq|jd kst|jd kst|jd kst|j\} } | d gkstt||||d\} } }t|j| t|j| t|j|dS) z:Check the default usage of the LearningCurveDisplay class.rrrrrrNg?Scorez%Number of samples in the training setTesting metric)rrr matplotlib errorbar_AssertionError isinstancelines_listlinesLine2D fill_between_ collectionsPolyCollectionZ get_alpha score_nameax_Z get_xlabel get_ylabelget_legend_handles_labelsrrrr train_scores test_scores)rr rrrrdisplaympllinefill_Z legend_labelstrain_sizes_absr1r2r r r )test_learning_curve_display_default_usage&s@       r9c Cs2|\}}tddd}dddg}d}tj|||||d}|jdd}|dksXt|jd ksjtd }tj|||||d}|jdd} | dkstt | | |jd kstd}tj|||||d}|jd kst|j | d |jd kst|jdddks.td S) zaCheck the behaviour of the `negate_score` parameter calling `from_estimator` and `plot`. rZ max_depthr rrrF)r negate_scorer T)r<N) rrrr&get_dataallr$r.r/rZplot) rr rrrrr<r3Zpositive_scoresZnegative_scoresr r r (test_learning_curve_display_negate_scoreMsL   r?zscore_name, ylabel)Nr )Accuracyr@c Cs|\}}tdd}dddg}tj|||||d}|j|ksBt|\}}tddd}dddg}tj|||||d}|j|kstd S) zGCheck that we can overwrite the default score name shown on the y-axis.rrrrr)rr-r:r;N)rrrr.r/r$r-) rr r-Zylabelrrrrr3r r r &test_learning_curve_display_score_namezs,    rA)NerrorbarcCs|\}}tdd}dddg}t||||d\}}} d} tj||||| |d} | j\} } | d gksht|d krt| jd kst| j d kst| jd \}}n8| jd kstt| j d kst| j dj d \}}t ||t ||jd d d } tj||||| |d} | j\} } | dgks0t|d krrt| jd ksNt| j d ks^t| jd \}}n<| jd kstt| j d kst| j dj d \}}t ||t || jd d d} tj||||| |d} | j\} } | d dgkst|d krXt| jdks"t| j d ks2t| jd \}}| jd  \}}nT| jd kshtt| j dks|t| j dj d \}}| j d j d \}}t ||t ||jd d t ||t || jd d d S)z:Check the behaviour of setting the `score_type` parameter.rrrrrrZtrain)rrr zTraining metricNr:)Zaxistestr!Zboth)rrrrr.r0r$lenr&r#r=r(rrZmean)rr r rrrrr8r1r2rr3r7 legend_labelZx_dataZy_dataZ x_data_trainZ y_data_trainZ x_data_testZ y_data_testr r r &test_learning_curve_display_score_types            rGcCs|\}}tdd}dddg}tj||||dd}|jdksBt|jd ksTttj||||d d}|jd kszt|jd kstd S) z1Check the behaviour of the parameter `log_scale`.rrrrrT)rZ log_scalelogZlinearFN)rrrr.Z get_xscaler$Z get_yscale)rr rrrrr3r r r %test_learning_curve_display_log_scales*  rIc Cs|\}}tdd}ddl}dddg}d}tj|||||d}t|jdksNtt|jd|jj sft|j dkstt|j dkst|j \} } t| dkstd }tj|||||d}t|jdkstt|jd|jj st|j dkstt|j dkstt|j d|jjst|j \} } t| dks>td }tj|||||d}|jdksftt|j dksztt|j d|jjst|j dkst|j \} } t| dkstdS) z9Check the behaviour of the parameter `std_display_style`.rrNrrr)rr r: fill_betweenrB)rr"rrrEr&r$r%r(r)r#r*r.r0r+r, containerZErrorbarContainer) rr rrrr4rr r3r7rFr r r -test_learning_curve_display_std_display_styles^  rLc Cs|\}}tdd}dddg}d}ddi}dd d }tj|||||||d } | jddks`tt| jdd d d d ggd }ddi} tj|||||| d} | j dj ddkstdS)zuCheck the behaviour of the different plotting keyword arguments: `line_kw`, `fill_between_kw`, and `errorbar_kw`.rrrrrrJcolorZredg?)rMalpha)rr line_kwfill_between_kwgrB)rr errorbar_kwN) rrrr&Z get_colorr$rr*Z get_facecolorr#r() rr rrrrr rOrPr3rQr r r 'test_learning_curve_display_plot_kwargs=s>       rR)rZsklearn.datasetsrZ sklearn.treerZ sklearn.utilsrZsklearn.utils._testingrrZsklearn.model_selectionrrZfixturer markZ parametrize ValueErrorrr9r?rArGrIrLrRr r r r s6         '-   Z: