\c@sddlmZddlZddljZddlZddlm Z m Z ddl m Z m Z ddl mZddlmZmZdd lmZdd lmZdd lmZmZmZdd lmZmZdd lmZddlmZddlm Z m!Z!ddlm"Z"ddl#m$Z$ddl%m&Z&ddl%m'Z'dddddgZ(dZ)de$j*e efdYZ+de$j*e e+efdYZ,dZ-e.dd d!e.d"Z/dS(#i(tprint_functionN(tABCMetatabstractmethodi(tlibsvmt liblinear(t libsvm_sparsei(t BaseEstimatortClassifierMixin(t LabelEncoder(t_ovr_decision_function(t check_arraytcheck_consistent_lengthtcheck_random_state(t column_or_1dt check_X_y(tcompute_class_weight(tsafe_sparse_dot(tcheck_is_fittedt_check_large_sparse(tcheck_classification_targets(tsix(tConvergenceWarning(tNotFittedErrortc_svctnu_svct one_classt epsilon_svrtnu_svrc Cs,|jdd}g}tjtjdg|g}xt|D]}|||||dddf}xt|d|D]}|||||dddf} ||d||||df} ||||||df} |jt| |t| | qWqEW|S(sGenerate primal coefficients from dual coefficients for the one-vs-one multi class LibSVM in the case of a linear kernel.iiN(tshapetnptcumsumthstacktrangetappendR( t dual_coeft n_supporttsupport_vectorstn_classtcoeftsv_locstclass1tsv1tclass2tsv2talpha1talpha2((s/lib/python2.7/site-packages/sklearn/svm/base.pyt_one_vs_one_coefs !((&"t BaseLibSVMcBseZdZdddddgZedZedZddZ d Z d Z d Z d Z d ZdZdZdZdZdZdZdZedZdZRS(sBase class for estimators that use libsvm as backing library This implements support vector machine classification and regression. Parameter documentation is in the derived `SVC` class. tlineartpolytrbftsigmoidt precomputedcCs|jtkr+tdt|jfn|dkrLd}t|n||_||_||_||_||_||_||_ ||_ | |_ | |_ | |_ | |_| |_||_||_dS(Ns&impl should be one of %s, %s was givenisXThe gamma value of 0.0 is invalid. Use 'auto' to set gamma to a value of 1 / n_features.(t_implt LIBSVM_IMPLt ValueErrortkerneltdegreetgammatcoef0ttoltCtnutepsilont shrinkingt probabilityt cache_sizet class_weighttverbosetmax_itert random_state(tselfR8R9R:R;R<R=R>R?R@RARBRCRDRERFtmsg((s/lib/python2.7/site-packages/sklearn/svm/base.pyt__init__Is*               cCs |jdkS(NR4(R8(RG((s/lib/python2.7/site-packages/sklearn/svm/base.pyt _pairwisegsc Cst|j}tj|}|rB|jdkrBtdn|oUt|j |_t||dt j dddddt \}}|j |}t j |dkrgn|dt j }tj|j}|d kr'|jd |jd kr'td d |jd |jd fn|jdkr_|jd |jd kr_tdn|jd d kr|jd |jd krtd|j|jfn|jdkr|r|j|j|jd }n |j}|jdkr4|d kr(d|jd ||_qd|_qt|j oP|jd k}|rt j|d rtjdtnd|jd |_n2|jdkrd|jd |_n |j|_|j} t| rd} n|jr|jn|j} |jrt dddn|j!t j"dj#} | ||||| d| |j|_$|j%j&|_'|j(|_)|jd!krt*|j+d kr|j%d9_%|j( |_(n|S("s5Fit the SVM model according to the given training data. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) 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, shape (n_samples,) Target values (class labels in classification, real numbers in regression) sample_weight : array-like, shape (n_samples,) Per-sample weights. Rescale C per sample. Higher weights force the classifier to put more emphasis on these points. Returns ------- self : object 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. R4s-Sparse precomputed kernels are not supported.tdtypetorderR=t accept_sparsetcsrtaccept_large_sparseiis"X and y have incompatible shapes. sX has %s samples, but y has %s.is(X.shape[0] should be equal to X.shape[1]ssample_weight and X have incompatible shapes: %r vs %r Note: Sparse matrices cannot be indexed w/boolean masks (use `indices=True` in CV).tscaletauto_deprecatedg?R0sThe default value of gamma will change from 'auto' to 'scale' in version 0.22 to account better for unscaled features. Set gamma explicitly to 'auto' or 'scale' to avoid this warning.tautos[LibSVM]tendttit random_seedRRiN(RPRQ(R0R4(RR(,R RFtspt isspmatrixR8t TypeErrortcallablet_sparseRRtfloat64tFalset_validate_targetstasarraytNoneR6tindexR5RR7R:tmultiplytmeantvart_gammatisclosetwarningstwarnt FutureWarningt _sparse_fitt _dense_fitRDtprinttrandinttiinfotmaxt shape_fit_t intercept_tcopyt _intercept_t dual_coef_t _dual_coef_tlentclasses_( RGtXtyt sample_weighttrndtsparset solver_typetX_vartkernel_uses_gammaR8tfittseed((s/lib/python2.7/site-packages/sklearn/svm/base.pyRlsl  &")-&             $cCs.tjd|_t|dtjtjS(sxValidation of y and class_weight. Default implementation for SVR and one-class; overridden in BaseSVC. iRh(Rtemptyt class_weight_R tTruetastypeR\(RGRy((s/lib/python2.7/site-packages/sklearn/svm/base.pyR^scCsB|jdkst|jdkr>tjd|jtndS(NiisnSolver terminated early (max_iter=%i). Consider pre-processing your data with StandardScaler or MinMaxScaler.(ii(t fit_status_tAssertionErrorRgRhRER(RG((s/lib/python2.7/site-packages/sklearn/svm/base.pyt_warn_from_fit_statuss c#Cs7t|jrS||_|j|}|jd|jdkrStdqSntj|jtj ||d|d|d|j d|d|j d |j d |j d |jd |jd |jd|jd|jd|jd|jd|jd|\|_|_|_|_|_|_|_|_|jdS(Niis(X.shape[0] should be equal to X.shape[1]tsvm_typeRzRCR8R=R>RAR9R@R<RBR;R:R?RERV(RZR8t_BaseLibSVM__Xfitt_compute_kernelRR7Rtset_verbosity_wrapRDRRR=R>RAR9R@R<RBR;ReR?REtsupport_tsupport_vectors_t n_support_RtRqtprobA_tprobB_RR(RGRxRyRzR}R8RV((s/lib/python2.7/site-packages/sklearn/svm/base.pyRks    Bc Cstj|jdtjdd|_|j|jj|}tj|j tj |j d|j|j |j ||||j|j|j|j|j|j||j|j|jt|jt|j|j|\|_|_}|_|_|_|_ |_!|j"t#|dr7t$|j%d} nd} |jj d} tj&tj'| | } tj'd| j(d| j(| } t)j*|| | f| | f|_+dS(NRKRLR=iRwi(,RR_tdataR\t sort_indicest_sparse_kernelsRaRRRDtlibsvm_sparse_trainRtindicestindptrR9ReR;R<R=RR>RBR?tintR@RARERRRqRRRRRthasattrRvRwttiletarangetsizeRWt csr_matrixRt( RGRxRyRzR}R8RVt kernel_typetdual_coef_dataR%tn_SVtdual_coef_indicestdual_coef_indptr((s/lib/python2.7/site-packages/sklearn/svm/base.pyRjs,$ " 6  cCs4|j|}|jr!|jn|j}||S(sPerform regression on samples in X. For an one-class model, +1 (inlier) or -1 (outlier) is returned. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) For kernel="precomputed", the expected shape of X is (n_samples_test, n_samples_train). Returns ------- y_pred : array, shape (n_samples,) (t_validate_for_predictR[t_sparse_predictt_dense_predict(RGRxtpredict((s/lib/python2.7/site-packages/sklearn/svm/base.pyR6scCs#|j\}}|j|}|jdkrHt|dddt}n|j}t|jrd}|jd|jdkrtd|jd|jdfqnt j |j }t j ||j|j|j|j|j|j|jd|d |d |jd |jd |jd |jS(NiRLR=ROR4isMX.shape[1] = %d should be equal to %d, the number of samples at training timeRR8R9R;R:RB(RRtndimR R]R8RZRpR7R6RaR5RRRRRRuRsRRR9R;ReRB(RGRxt n_samplest n_featuresR8R((s/lib/python2.7/site-packages/sklearn/svm/base.pyRIs" $ cCs|j}t|rd}n|jj|}d}tj|j|j|j|j j|j j|j j|j j|j t j|j ||j|j|j|j||j|j|j|j|j|j|j|jS(NR4g(R8RZRRaRtlibsvm_sparse_predictRRRRRuRsR6R5R9ReR;R<RR>R?R@RARRR(RGRxR8RR=((s/lib/python2.7/site-packages/sklearn/svm/base.pyR`s"        cCsgt|jrc|j||j}tj|rB|j}ntj|dtjdd}n|S(s0Return the data transformed by a callable kernelRKRLR=( RZR8RRWtissparsettoarrayRR_R\(RGRxR8((s/lib/python2.7/site-packages/sklearn/svm/base.pyRws !cCs{|j|}|j|}|jr9|j|}n|j|}|jdkrwt|jdkrw|j S|S(sbEvaluates the decision function for the samples in X. Parameters ---------- X : array-like, shape (n_samples, n_features) Returns ------- X : array-like, shape (n_samples, n_class * (n_class-1) / 2) Returns the decision function of the sample for each class in the model. RRi(RR( RRR[t_sparse_decision_functiont_dense_decision_functionR5RvRwtravel(RGRxtdec_func((s/lib/python2.7/site-packages/sklearn/svm/base.pyt_decision_functions $ cCst|dtjdddt}|j}t|r?d}ntj||j|j |j |j |j |j |jdtj|jd|d|jd |jd |jd |jS( NRKRLR=ROR4RR8R9RBR;R:(R RR\R]R8RZRtdecision_functionRRRRuRsRRR6RaR5R9RBR;Re(RGRxR8((s/lib/python2.7/site-packages/sklearn/svm/base.pyRs     cCstj|jdtjdd|_|j}t|drEd}n|jj|}tj |j|j |j |j j|j j |j j |j j|jtj|j||j|j|j|j|j|j|j|j|j|j|j|j|jS(NRKRLR=t__call__R4(RR_RR\R8RRRaRtlibsvm_sparse_decision_functionRRRRuRsR6R5R9ReR;R<R=RR>R?R@RARRR(RGRxR8R((s/lib/python2.7/site-packages/sklearn/svm/base.pyRs"$       c CsSt|dt|dddtjdddt}|jr_tj| r_tj|}n|jru|j ntj |r|j rt |j  rt dt|jn|j\}}|j d kr|jd |jd krOt d |jd |jd fqOn3||jd krOt d ||jd fn|S(NRRMRNRKRLR=ROs3cannot use sparse input in %r trained on dense dataR4iisMX.shape[1] = %d should be equal to %d, the number of samples at training timesNX.shape[1] = %d should be equal to %d, the number of features at training time(RR RR\R]R[RWRXRRRRZR8R7ttypet__name__RRp(RGRxRR((s/lib/python2.7/site-packages/sklearn/svm/base.pyRs(    )$cCs[|jdkrtdn|j}tj|rKt|jj_n t|j_|S(NR0s2coef_ is only available when using a linear kernel( R8tAttributeErrort _get_coefRWRR]Rtflagst writeable(RGR&((s/lib/python2.7/site-packages/sklearn/svm/base.pytcoef_s  cCst|j|jS(N(RRuR(RG((s/lib/python2.7/site-packages/sklearn/svm/base.pyRsN(Rt __module__t__doc__RRRItpropertyRJR`RR^RRkRjRRRRRRRRRR(((s/lib/python2.7/site-packages/sklearn/svm/base.pyR/<s& x   #       tBaseSVCcBseZdZedZdZdZdZdZe dZ dZ e dZ d Z d Zd Zd ZRS( s!ABC for LibSVM-based classifiers.cCsz||_tt|jd|d|d|d|d|d|d|dd d |d | d | d | d| d| d|dS(NR8R9R:R;R<R=R>R?gR@RARBRCRDRERF(tdecision_function_shapetsuperRRI(RGR8R9R:R;R<R=R>R@RARBRCRDRERRF((s/lib/python2.7/site-packages/sklearn/svm/base.pyRIs $ cCst|dt}t|tj|dt\}}t|j|||_t|dkrzt dt|n||_ tj |dtj ddS(NRhtreturn_inverseis>The number of classes has to be greater than one; got %d classRKRLR=( R RRRtuniqueRRCRRvR7RwR_R\(RGRyty_tcls((s/lib/python2.7/site-packages/sklearn/svm/base.pyR^s  cCsW|j|}|jdkrSt|jdkrSt|dk| t|jS|S(sEvaluates the decision function for the samples in X. Parameters ---------- X : array-like, shape (n_samples, n_features) Returns ------- X : array-like, 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. tovrii(RRRvRwR (RGRxtdec((s/lib/python2.7/site-packages/sklearn/svm/base.pyRs$ cCs:tt|j|}|jjtj|dtjS(sPerform classification on samples in X. For an one-class model, +1 or -1 is returned. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) For kernel="precomputed", the expected shape of X is [n_samples_test, n_samples_train] Returns ------- y_pred : array, shape (n_samples,) Class labels for samples in X. RK(RRRRwttakeRR_tintp(RGRxRy((s/lib/python2.7/site-packages/sklearn/svm/base.pyR0scCs:|jstdn|jdkr6tdndS(Ns6predict_proba is not available when probability=FalseRRs0predict_proba only implemented for SVC and NuSVC(RR(RARR5(RG((s/lib/python2.7/site-packages/sklearn/svm/base.pyt _check_probaGs cCs|j|jS(sCompute 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, shape (n_samples, n_features) For kernel="precomputed", the expected shape of X is [n_samples_test, n_samples_train] Returns ------- T : array-like, 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 `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. (Rt_predict_proba(RG((s/lib/python2.7/site-packages/sklearn/svm/base.pyt predict_probaOs cCsg|j|}|jjdks3|jjdkrBtdn|jrT|jn|j}||S(NisApredict_proba is not available when fitted with probability=False(RRRRRR[t_sparse_predict_probat_dense_predict_proba(RGRxt pred_proba((s/lib/python2.7/site-packages/sklearn/svm/base.pyRms $cCs|j|jS(sCompute log probabilities of possible outcomes for samples in X. The model need to have probability information computed at training time: fit with attribute `probability` set to True. Parameters ---------- X : array-like, shape (n_samples, n_features) For kernel="precomputed", the expected shape of X is [n_samples_test, n_samples_train] Returns ------- T : array-like, shape (n_samples, n_classes) Returns the log-probabilities of the sample for each class in the model. The columns correspond to the classes in sorted order, as they appear in the attribute `classes_`. Notes ----- The probability model is created using cross validation, so the results can be slightly different than those obtained by predict. Also, it will produce meaningless results on very small datasets. (Rt_predict_log_proba(RG((s/lib/python2.7/site-packages/sklearn/svm/base.pytpredict_log_probavs cCstj|j|S(N(RtlogR(RGRx((s/lib/python2.7/site-packages/sklearn/svm/base.pyRscCs|j|}|j}t|r-d}ntj|j}tj||j|j |j |j |j |j |jd|d|d|jd|jd|jd|j}|S(NR4RR8R9RBR;R:(RR8RZR6RaR5RRRRRRuRsRRR9RBR;Re(RGRxR8Rtpprob((s/lib/python2.7/site-packages/sklearn/svm/base.pyRs    cCstj|jdtjdd|_|j}t|rBd}n|jj|}tj |j|j |j |j j|j j |j j |j j|jtj|j||j|j|j|j|j|j|j|j|j|j|j|j|jS(NRKRLR=R4(RR_RR\R8RZRRaRtlibsvm_sparse_predict_probaRRRRuRsR6R5R9ReR;R<R=RR>R?R@RARRR(RGRxR8R((s/lib/python2.7/site-packages/sklearn/svm/base.pyRs"$        cCs|jjddkr.t|j|j}nUt|j|j|j}tj|drttj|j }nt j|}|S(Nii( RtRRRR.RRWRtvstackttocsrR(RGR&((s/lib/python2.7/site-packages/sklearn/svm/base.pyRs (RRRRRIR^RRRRRRRRRRR(((s/lib/python2.7/site-packages/sklearn/svm/base.pyRs        c Csiiidt6d6idt6dt6d6d6iidt6d6d6iid t6d6id t6d t6d6d 6iid t6d6d6iidt6dt6d6d6dd6}|dkr||S|dkrtd|n|j|d}|dkrd|}ni|j|d}|dkr6d||f}n8|j|d}|dkrjd|||f}n|Std||||fdS(sFind 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. itl1iitl2tlogistic_regressionithingeiiit squared_hingei tepsilon_insensitivei i tsquared_epsilon_insensitiveitcrammer_singerRs<`multi_class` must be one of `ovr`, `crammer_singer`, got %rsloss='%s' is not supporteds>The combination of penalty='%s' and loss='%s' is not supportedsLThe combination of penalty='%s' and loss='%s' are not supported when dual=%ssJUnsupported set of arguments: %s, Parameters: penalty=%r, loss=%r, dual=%rN(R]RR7tgetR`( t multi_classtpenaltytlosstdualt_solver_type_dictt _solver_pent error_stringt _solver_dualt solver_num((s/lib/python2.7/site-packages/sklearn/svm/base.pyt_get_liblinear_solver_typesB        RRg?c Cs| dkrkt}|j|}|j}t|dkrVtd|dnt|||}ntjddtj}|}t j |t | }|rt ddd nd }|r|dkrtd |q|}nt j |tj |t j |tj|r2t|ntj|dtjj}tj|d d }|dkrtj|jd}n+tj|dtjdd}t||t| || |}t j||tj||| |||| |jtjdj|| \}}t|}|| krIt j!dt"n|r|ddddf}||dddf}n |}d}|||fS(s 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}, shape (n_samples, n_features) Training vector, where n_samples in the number of samples and n_features is the number of features. y : array-like, 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, 'balanced'}, optional 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 : str, {'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, optional (default=None) The seed of the pseudo random number generator to use when shuffling the data. If int, random_state is the seed used by the random number generator; If RandomState instance, random_state is the random number generator; If None, the random number generator is the RandomState instance used by `np.random`. multi_class : str, {'ovr', 'crammer_singer'} `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 : str, {'logistic_regression', 'hinge', 'squared_hinge', 'epsilon_insensitive', 'squared_epsilon_insensitive} The loss function used to fit the model. epsilon : float, optional (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, optional Weights assigned to each sample. Returns ------- coef_ : ndarray, 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_ : int Maximum number of iterations run across all classes. RRiseThis solver needs samples of at least 2 classes in the data, but the data contains only one class: %riRKs [LibLinear]RSRTgsqIntercept scaling is %r but needs to be greater than 0. To disable fitting an intercept, set fit_intercept=False.t requirementstWRLR=RUs@Liblinear failed to converge, increase the number of iterations.Nig(RR(#Rt fit_transformRwRvR7RRRR\RRR RlRRRWRRR_RtrequireR`tonesRtarrayR Rt train_wrapRXRmRnRoRgRhR(RxRyR=t fit_intercepttintercept_scalingRCRRRDRER<RFRRR?Rztencty_indRwRR{tbiasR}t raw_coef_tn_iter_RRq((s/lib/python2.7/site-packages/sklearn/svm/base.pyt_fit_liblinear s\`                  (0t __future__RtnumpyRt scipy.sparseR|RWRgtabcRRRTRRRtbaseRRt preprocessingRtutils.multiclassR tutilsR R R R RRt utils.extmathRtutils.validationRRRt externalsRt exceptionsRRR6R.twith_metaclassR/RRR`R(((s/lib/python2.7/site-packages/sklearn/svm/base.pyts6   !"% >