U 3d@s~ddlZddlmZmZddlmZmZddlZddl m Z ddl m Zddl mZddl mZdd lmZmZdd lmZdd lmZdd lmZmZdd lmZddlmZddlm Z ddl!m"Z"ddl#m$Z$m%Z%ddl#m&Z&ddl#m'Z'm(Z(ddlm)Z)ddl*m+Z+m,Z,ddl-m.Z.ddl-m/Z/dddddgZ0ddZ1Gdd d eed!Z2Gd"d#d#ee2ed!Z3d$d%Z4d+d)d*Z5dS),N)ABCMetaabstractmethod)IntegralReal)_libsvm) _liblinear)_libsvm_sparse) BaseEstimatorClassifierMixin) LabelEncoder)_ovr_decision_function) check_arraycheck_random_state) column_or_1d)compute_class_weight) available_if)safe_sparse_dot)check_is_fitted_check_large_sparse) _num_samples)_check_sample_weightcheck_consistent_length)check_classification_targets)Interval StrOptions)ConvergenceWarning)NotFittedErrorc_svcnu_svcZ one_classZ epsilon_svrZnu_svrc Cs|jdd}g}ttdg|g}t|D]}|||||dddf}t|d|D]z}|||||dddf} ||d||||df} ||||||df} |t| |t| | qbq0|S)zGenerate primal coefficients from dual coefficients for the one-vs-one multi class LibSVM in the case of a linear kernel.rrN)shapenpZcumsumZhstackrangeappendr) dual_coefZ n_supportZsupport_vectorsn_classcoefZsv_locsZclass1Zsv1Zclass2Zsv2Zalpha1Zalpha2r(5/tmp/pip-unpacked-wheel-zrfo1fqw/sklearn/svm/_base.py_one_vs_one_coef!s     r*c@seZdZUdZedddddhegeeddd d ged d heed dd d geedddd geed ddd geed ddd geed ddd geed dd d gdgdgeedddd gedhe dgdgeeddd d gdgdZ e e d<dddddgZ e ddZddZdr?r@rArBrCrDrEr6rFr8)selfr:r;r<r=r>r?r@rArBrCrDrEr6rFr8r(r(r)__init__fs&  zBaseLibSVM.__init__cCsd|jdkiS)NZpairwiser-)r:rKr(r(r) _more_tagsszBaseLibSVM._more_tagscCs,|t|j}t|}|r2|jdkr2td|o@t|j |_t|jrZt ||n|j ||t j dddd\}}| |}t j|dkrgn|t j d}t|j}t|}|d kr||jd krtd d ||jd f|jdkr||jd krtd|jd |jd |jd d krL|jd |krLtd|j|jft|jr\dn|j}|dkrtd|_nt|jtr|jdkr|r|||d n|} | d krd|jd | nd|_n|jdkrd|jd |_nt|jtr|j|_|jr|jn|j} |j r6t!ddd|"t #dj$} | |||||| dt%|drn|jn|f|_&|j'(|_)|j*|_+|jdkrt,|j-d kr|j'd9_'|j* |_*|jr|j+j.n|j+} t /|j)0} t /| 0}| r|std|jdkr|j1|_2n |j13|_2|S)aFit the SVM model according to the given training data. Parameters ---------- X : {array-like, sparse matrix} of shape (n_samples, n_features) or (n_samples, n_samples) Training vectors, where `n_samples` is the number of samples and `n_features` is the number of features. For kernel="precomputed", the expected shape of X is (n_samples, n_samples). y : array-like of shape (n_samples,) Target values (class labels in classification, real numbers in regression). sample_weight : array-like of shape (n_samples,), default=None Per-sample weights. Rescale C per sample. Higher weights force the classifier to put more emphasis on these points. Returns ------- self : object Fitted estimator. Notes ----- If X and y are not C-ordered and contiguous arrays of np.float64 and X is not a scipy.sparse.csr_matrix, X and/or y may be copied. If X is a dense array, then the other methods will not support sparse matrices as input. r-z-Sparse precomputed kernels are not supported.r?csrF)dtypeorder accept_sparseaccept_large_sparseNrPr rz"X and y have incompatible shapes. zX has %s samples, but y has %s.rzDPrecomputed matrix must be a square matrix. Input is a {}x{} matrix.zsample_weight and X have incompatible shapes: %r vs %r Note: Sparse matrices cannot be indexed w/boolean masks (use `indices=True` in CV).r2r0r3r1z[LibSVM]endi) random_seedr!rr r7zwThe dual coefficients or intercepts are not finite. The input data may contain large values and need to bepreprocessed.)4Z_validate_paramsrr8sp isspmatrixr: TypeErrorcallable_sparser_validate_datar"float64_validate_targetsasarrayrIindexrHrr!rJformat_gamma isinstancer<strmultiplyZmeanvarr _sparse_fit _dense_fitr6printrandintiinfomaxhasattr shape_fit_ intercept_copy _intercept_ dual_coef_ _dual_coef_lenclasses_dataisfiniteall _num_itern_iter_item)rKXy sample_weightrndsparse solver_typeZ n_samplesr:ZX_varfitseedr%Zintercept_finitenessZdual_coef_finitenessr(r(r)rs!          ($        zBaseLibSVM.fitcCst|ddjtjddS)zxValidation of y and class_weight. Default implementation for SVR and one-class; overridden in BaseSVC. TwarnF)rt)rZastyper"ra)rKrr(r(r)rb!szBaseLibSVM._validate_targetscCs.|jdkst|jdkr*td|jtdS)NrrrznSolver terminated early (max_iter=%i). Consider pre-processing your data with StandardScaler or MinMaxScaler.) fit_status_AssertionErrorwarningsrrFrrMr(r(r)_warn_from_fit_status(s z BaseLibSVM._warn_from_fit_statuscCst|jr6||_||}|jd|jdkr6tdt|jtj ||||t |dt d||j |j|j|j|j|j|j|j|j|j|j|d\ |_|_|_|_|_|_|_|_|_ |!dS)Nrrz(X.shape[0] should be equal to X.shape[1] _class_weight)svm_typerrEr:r?r@rCr;rBr>rDr=r<rArFrY)"r^r:_BaseLibSVM__Xfit_compute_kernelr!rJlibsvmset_verbosity_wrapr6rgetattrr"emptyr?r@rCr;rBr>rDr=rfrArFsupport_support_vectors_ _n_supportrvrs_probA_probBrr}r)rKrrrrr:rYr(r(r)rl3sJ   zBaseLibSVM._dense_fitc CsPtj|jtjdd|_||j|}t|j t |j d|j|j |j ||||j|j|j|j|jt|dtd||j|j|jt|jt|j|j|\ |_|_}|_|_|_ |_!|_"|_#|$t%|drt&|j'd} nd} |jj d} t(t)| | } | st*+g|_,n2t)d| j-d| j-| } t*+|| | f| | f|_,dS)Nr?rPrQrrrry).r"rcrzra sort_indices_sparse_kernelsrd libsvm_sparserr6Zlibsvm_sparse_trainr!indicesindptrr;rfr=r>r?rrr@rDrAintrBrCrFrrrsrrrrr}rrqrxryZtileZaranger[ csr_matrixrvsize) rKrrrrr:rY kernel_typeZdual_coef_datar&Zn_SVZdual_coef_indicesZdual_coef_indptrr(r(r)rkcsl    zBaseLibSVM._sparse_fitcCs$||}|jr|jn|j}||S)aPerform regression on samples in X. For an one-class model, +1 (inlier) or -1 (outlier) is returned. Parameters ---------- X : {array-like, sparse matrix} of shape (n_samples, n_features) For kernel="precomputed", the expected shape of X is (n_samples_test, n_samples_train). Returns ------- y_pred : ndarray of shape (n_samples,) The predicted values. )_validate_for_predictr__sparse_predict_dense_predict)rKrpredictr(r(r)rs zBaseLibSVM.predictcCs||}|jdkr"t|ddd}|j}t|jrfd}|jd|jdkrftd|jd|jdft |j }t j ||j |j|j|j|j|j|j|||j|j|j|jdS) Nrr?F)rQrSr-rMX.shape[1] = %d should be equal to %d, the number of samples at training time)rr:r;r=r<rD)rndimrr:r^r!rrrJrIrdrHrrrrrrwrurrr;r=rfrD)rKrr:rr(r(r)rs:    zBaseLibSVM._dense_predictcCs|j}t|rd}|j|}d}t|j|j|j|j j|j j|j j|j j|j t |j ||j|j|j|j|t|dtd|j|j|j|j|j|j|jS)Nr-r2rr)r:r^rrdrZlibsvm_sparse_predictrzrrrrwrurIrHr;rfr=r>rr"rr@rArBrCrrr)rKrr:rr?r(r(r)rs<  zBaseLibSVM._sparse_predictcCs@t|jr<|||j}t|r*|}tj|tjdd}|S)z0Return the data transformed by a callable kernelr?r) r^r:rr[issparseZtoarrayr"rcrarKrr:r(r(r)rs   zBaseLibSVM._compute_kernelcCsV||}||}|jr&||}n ||}|jdkrRt|jdkrR| S|S)afEvaluates the decision function for the samples in X. Parameters ---------- X : array-like of shape (n_samples, n_features) Returns ------- X : array-like of shape (n_samples, n_class * (n_class-1) / 2) Returns the decision function of the sample for each class in the model. rZr ) rrr__sparse_decision_function_dense_decision_functionrHrxryravel)rKrZdec_funcr(r(r)_decision_functions     zBaseLibSVM._decision_functioncCsht|tjddd}|j}t|r$d}tj||j|j|j |j |j |j |j t|j||j|j|j|jdS)Nr?F)rPrQrSr-rr:r;rDr=r<)rr"rar:r^rdecision_functionrrrrwrurrrIrdrHr;rDr=rfrr(r(r)r%s( z#BaseLibSVM._dense_decision_functioncCstj|jtjdd|_|j}t|dr*d}|j|}t |j|j |j |j j|j j |j j |j j|jt|j||j|j|j|j|jt|dtd|j|j|j|j|j|j|jS)Nr?r__call__r-rr)r"rcrzrar:rqrrdrZlibsvm_sparse_decision_functionrrrrwrurIrHr;rfr=r>r?rrr@rArBrCrrrrKrr:rr(r(r)r=s<   z$BaseLibSVM._sparse_decision_functioncCst|t|js*|j|dtjdddd}|jrDt|sDt |}|jrR| t |r~|js~t|js~t dt |j|jdkr|jd|jdkrt d |jd|jdf|j}|js|jdkr|j|jdkrt d |jjd |S) NrOr?F)rRrPrQrSresetz3cannot use sparse input in %r trained on dense datar-rrrzThe internal representation of z was altered)rr^r:r`r"rar_r[r\rrrrJtype__name__r!rrrr n_support_sum __class__)rKrsvr(r(r)rasB    $z BaseLibSVM._validate_for_predictcCs<|jdkrtd|}t|r0d|jj_nd|j_|S)zWeights assigned to the features when `kernel="linear"`. Returns ------- ndarray of shape (n_features, n_classes) r,z2coef_ is only available when using a linear kernelF)r:AttributeError _get_coefr[rrzflagsZ writeablerKr'r(r(r)coef_s   zBaseLibSVM.coef_cCst|j|jS)N)rrwrrMr(r(r)rszBaseLibSVM._get_coefcCsVz t|Wntk r$tYnXt|j}|dkr@|jSt|jdgSdS)z)Number of support vectors for each class.rrN) rrrrIrdrHrr"array)rKrr(r(r)rs   zBaseLibSVM.n_support_)N)!r __module__ __qualname____doc__rr^rrrdictrG__annotations__rrrLrNrrbrrlrkrrrrrrrrpropertyrrrr(r(r(r)r+AsX    (  0>"% $( r+) metaclasscseZdZUdZejeddhgdgdZeed<dD]Z e e q6e fdd Z d d Z d d ZfddZddZeeddZeeddZddZddZddZeddZeddZed d!ZZS)"BaseSVCz!ABC for LibSVM-based classifiers.ovrovor5)decision_function_shape break_tiesrG)rAr@cs:||_||_tj|||||||d|| | | | | |ddS)Nr2r9)rrsuperrL)rKr:r;r<r=r>r?r@rBrCrDrEr6rFrr8rrr(r)rLs&zBaseSVC.__init__cCslt|dd}t|tj|dd\}}t|j||d|_t|dkrTtdt|||_ tj |tj ddS) NTr)Zreturn_inverseclassesrr z>The number of classes has to be greater than one; got %d classr?r) rrr"uniquerrE class_weight_rxrJryrcra)rKrZy_clsr(r(r)rbs  zBaseSVC._validate_targetscCs>||}|jdkr:t|jdkr:t|dk| t|jS|S)a4Evaluate the decision function for the samples in X. Parameters ---------- X : array-like of shape (n_samples, n_features) The input samples. Returns ------- X : ndarray of shape (n_samples, n_classes * (n_classes-1) / 2) Returns the decision function of the sample for each class in the model. If decision_function_shape='ovr', the shape is (n_samples, n_classes). Notes ----- If decision_function_shape='ovo', the function values are proportional to the distance of the samples X to the separating hyperplane. If the exact distances are required, divide the function values by the norm of the weight vector (``coef_``). See also `this question `_ for further details. If decision_function_shape='ovr', the decision function is a monotonic transformation of ovo decision function. rr r)rrrxryr)rKrdecr(r(r)rs zBaseSVC.decision_functioncsxt||jr |jdkr td|jrT|jdkrTt|jdkrTtj||dd}n t |}|j tj |tj dS)aPerform classification on samples in X. For an one-class model, +1 or -1 is returned. Parameters ---------- X : {array-like, sparse matrix} of shape (n_samples, n_features) or (n_samples_test, n_samples_train) For kernel="precomputed", the expected shape of X is (n_samples_test, n_samples_train). Returns ------- y_pred : ndarray of shape (n_samples,) Class labels for samples in X. rz>break_ties must be False when decision_function_shape is 'ovo'rr r)ZaxisrT)rrrrJrxryr"ZargmaxrrrZtakercZintp)rKrrrr(r)rs  zBaseSVC.predictcCs$|jstd|jdkr tddS)Nz6predict_proba is not available when probability=FalserZz0predict_proba only implemented for SVC and NuSVCT)rCrrHrMr(r(r) _check_proba;s zBaseSVC._check_probacCsD||}|jjdks"|jjdkr*td|jr6|jn|j}||S)aCompute probabilities of possible outcomes for samples in X. The model need to have probability information computed at training time: fit with attribute `probability` set to True. Parameters ---------- X : array-like of shape (n_samples, n_features) For kernel="precomputed", the expected shape of X is (n_samples_test, n_samples_train). Returns ------- T : ndarray of shape (n_samples, n_classes) Returns the probability of the sample for each class in the model. The columns correspond to the classes in sorted order, as they appear in the attribute :term:`classes_`. Notes ----- The probability model is created using cross validation, so the results can be slightly different than those obtained by predict. Also, it will produce meaningless results on very small datasets. rzApredict_proba is not available when fitted with probability=False)rprobA_rprobB_rr__sparse_predict_proba_dense_predict_proba)rKrZ pred_probar(r(r) predict_probaDs zBaseSVC.predict_probacCst||S)aCompute log probabilities of possible outcomes for samples in X. The model need to have probability information computed at training time: fit with attribute `probability` set to True. Parameters ---------- X : array-like of shape (n_samples, n_features) or (n_samples_test, n_samples_train) For kernel="precomputed", the expected shape of X is (n_samples_test, n_samples_train). Returns ------- T : ndarray of shape (n_samples, n_classes) Returns the log-probabilities of the sample for each class in the model. The columns correspond to the classes in sorted order, as they appear in the attribute :term:`classes_`. Notes ----- The probability model is created using cross validation, so the results can be slightly different than those obtained by predict. Also, it will produce meaningless results on very small datasets. )r"logr)rKrr(r(r)predict_log_probaiszBaseSVC.predict_log_probacCsh||}|j}t|rd}t|j}tj||j|j |j |j |j |j |j|||j|j|j|jd}|S)Nr-r)rr:r^rIrdrHrrrrrrwrurrr;rDr=rf)rKrr:rZpprobr(r(r)rs,  zBaseSVC._dense_predict_probacCstj|jtjdd|_|j}t|r(d}|j|}t |j|j |j |j j|j j |j j |j j|jt|j||j|j|j|j|jt|dtd|j|j|j|j|j|j|jS)Nr?rr-rr)r"rcrzrar:r^rrdrZlibsvm_sparse_predict_probarrrrwrurIrHr;rfr=r>r?rrr@rArBrCrrrrr(r(r)rs<  zBaseSVC._sparse_predict_probacCs^|jjddkr t|j|j}n:t|j|j|j}t|drPt| }n t |}|S)Nrr) rvr!rrr*rr[rZvstackZtocsrr"rr(r(r)rs zBaseSVC._get_coefcCs|jSzParameter learned in Platt scaling when `probability=True`. Returns ------- ndarray of shape (n_classes * (n_classes - 1) / 2) )rrMr(r(r)rszBaseSVC.probA_cCs|jSr)rrMr(r(r)rszBaseSVC.probB_cCs|jS)zWeights per class)rrMr(r(r)rszBaseSVC._class_weight)rrrrr+rGrrrZ unused_parampoprrLrbrrrrrrrrrrrrr __classcell__r(r(rr)rs4   ' %  $ $  rc Csddiddddddd iidd id d ddddd iiddddidd}|dkr\||S|dkrptd|||d}|dkrd|}nJ||d}|dkrd||f}n(||d}|dkrd|||f}n|Std||||fdS)aFind the liblinear magic number for the solver. This number depends on the values of the following attributes: - multi_class - penalty - loss - dual The same number is also internally used by LibLinear to determine which solver to use. Fr)FT)l1l2rTr r )logistic_regressionZhingeZ squared_hingeepsilon_insensitivesquared_epsilon_insensitivecrammer_singerrrz<`multi_class` must be one of `ovr`, `crammer_singer`, got %rNzloss='%s' is not supportedz>The combination of penalty='%s' and loss='%s' is not supportedzLThe combination of penalty='%s' and loss='%s' are not supported when dual=%szJUnsupported set of arguments: %s, Parameters: penalty=%r, loss=%r, dual=%r)rJget) multi_classpenaltylossdualZ_solver_type_dictZ _solver_penZ error_stringZ _solver_dualZ solver_numr(r(r)_get_liblinear_solver_typesD         rrr皙?cCs| dkrJt}||}|j}t|dkr:td|dt|||d}ntjdtjd}|}t |t | }|rt ddd d }|r|dkrtd |n|}t |t |t |t|rt|tj|tjd}tj|d d }t||tjd}t| || |}t ||t||| |||| |tdj|| \}}t|}|| krhtdt|r|ddddf}||dddf}n|}d}|||fS)a Used by Logistic Regression (and CV) and LinearSVC/LinearSVR. Preprocessing is done in this function before supplying it to liblinear. Parameters ---------- X : {array-like, sparse matrix} of shape (n_samples, n_features) Training vector, where `n_samples` is the number of samples and `n_features` is the number of features. y : array-like of shape (n_samples,) Target vector relative to X C : float Inverse of cross-validation parameter. Lower the C, the more the penalization. fit_intercept : bool Whether or not to fit the intercept, that is to add a intercept term to the decision function. intercept_scaling : float LibLinear internally penalizes the intercept and this term is subject to regularization just like the other terms of the feature vector. In order to avoid this, one should increase the intercept_scaling. such that the feature vector becomes [x, intercept_scaling]. class_weight : dict or 'balanced', default=None Weights associated with classes in the form ``{class_label: weight}``. If not given, all classes are supposed to have weight one. For multi-output problems, a list of dicts can be provided in the same order as the columns of y. The "balanced" mode uses the values of y to automatically adjust weights inversely proportional to class frequencies in the input data as ``n_samples / (n_classes * np.bincount(y))`` penalty : {'l1', 'l2'} The norm of the penalty used in regularization. dual : bool Dual or primal formulation, verbose : int Set verbose to any positive number for verbosity. max_iter : int Number of iterations. tol : float Stopping condition. random_state : int, RandomState instance or None, default=None Controls the pseudo random number generation for shuffling the data. Pass an int for reproducible output across multiple function calls. See :term:`Glossary `. multi_class : {'ovr', 'crammer_singer'}, default='ovr' `ovr` trains n_classes one-vs-rest classifiers, while `crammer_singer` optimizes a joint objective over all classes. While `crammer_singer` is interesting from an theoretical perspective as it is consistent it is seldom used in practice and rarely leads to better accuracy and is more expensive to compute. If `crammer_singer` is chosen, the options loss, penalty and dual will be ignored. loss : {'logistic_regression', 'hinge', 'squared_hinge', 'epsilon_insensitive', 'squared_epsilon_insensitive}, default='logistic_regression' The loss function used to fit the model. epsilon : float, default=0.1 Epsilon parameter in the epsilon-insensitive loss function. Note that the value of this parameter depends on the scale of the target variable y. If unsure, set epsilon=0. sample_weight : array-like of shape (n_samples,), default=None Weights assigned to each sample. Returns ------- coef_ : ndarray of shape (n_features, n_features + 1) The coefficient vector got by minimizing the objective function. intercept_ : float The intercept term added to the vector. n_iter_ : array of int Number of iterations run across for each class. )rrr zeThis solver needs samples of at least 2 classes in the data, but the data contains only one class: %rrrrTz [LibLinear]rUrVgzqIntercept scaling is %r but needs to be greater than 0. To disable fitting an intercept, set fit_intercept=False.W) requirementsrXz@Liblinear failed to converge, increase the number of iterations.Nr7r2)r Z fit_transformryrxrJrr"rra liblinearrrrmrrr[rrrcrrequirerrZ train_wrapr\rnrorprrr)rrr?Z fit_interceptZintercept_scalingrErrr6rFr>r8rrrArencZy_indryrrZbiasrZ raw_coef_r~Z n_iter_maxrrsr(r(r)_fit_liblinear,szl         r)NrrrN)6rabcrrZnumbersrrZnumpyr"Z scipy.sparserr[rUrrrrr rbaser r Z preprocessingr Zutils.multiclassrutilsrrrrZutils.metaestimatorsrZ utils.extmathrZutils.validationrrrrrrZutils._param_validationrr exceptionsrrrIr*r+rrrr(r(r(r)sN               wAE