0PhgrdZddlZddlmZmZddlmZmZddlZ ddl m Z m Z m Z ddlmZmZmZmZmZmZddlmZdd lmZmZdd lmZmZdd lmZmZm Z dd l!m"Z"dd l#m$Z$ddl%m&Z&ddl'm(Z(m)Z)ddl*m+Z+m,Z,m-Z-ddl.m/Z/m0Z0m1Z1ddl2m3Z3m4Z4m5Z5m6Z6m7Z7m8Z8m9Z9dddddddZ:dddddZ;dZ<e j=e j>j?Z@GddZAGdde0eeZBd.d!ZC d/d"ZDd#ZEGd$d%e/eBeZFGd&d'eFZGGd(d)eeBZHGd*d+eHZIGd,d-eeBZJdS)0zVClassification, regression and One-Class SVM using Stochastic Gradient Descent (SGD). N)ABCMetaabstractmethod)IntegralReal)CyHalfBinomialLossCyHalfSquaredError CyHuberLoss) BaseEstimator OutlierMixinRegressorMixin _fit_contextclone is_classifier)ConvergenceWarning) ShuffleSplitStratifiedShuffleSplit)check_random_statecompute_class_weight)HiddenInterval StrOptions)safe_sparse_dot) available_if)_check_partial_fit_first_call)Paralleldelayed)_check_sample_weightcheck_is_fitted validate_data)LinearClassifierMixinSparseCoefMixin make_dataset)EpsilonInsensitiveHinge ModifiedHuberSquaredEpsilonInsensitive SquaredHinge _plain_sgd32 _plain_sgd64)constantoptimal invscalingadaptivepa1pa2)nonel2l1 elasticnet皙?c eZdZdZddZdZdS)_ValidationScoreCallbackz5Callback for early stopping based on validation scoreNct||_d|j_| ||j_||_||_||_dS)Nr!)r estimatort_classes_X_valy_valsample_weight_val)selfr>rArBrCclassess i/var/www/html/test/jupyter/venv/lib/python3.11/site-packages/sklearn/linear_model/_stochastic_gradient.py__init__z!_ValidationScoreCallback.__init__?sGy))  &-DN #  !2c|j}|dd|_tj||_||j|j|j S)Nr!) r>reshapecoef_np atleast_1d intercept_scorerArBrC)rDcoef interceptests rF__call__z!_ValidationScoreCallback.__call__HsLnLLB'' y11yyTZ1GHHHrHN)__name__ __module__ __qualname____doc__rGrTrHrFr<r<<sB??3333IIIIIrHr<ceZdZUdZdgeedddgeeddddgdgdgd gdgeeddddgd Zee d <d d dddddddddddddddddddZ e dZ d#dZ dZdZdZ d$d Zd!Z d%d"ZdS)&BaseSGDz1Base class for SGD classification and regression.booleanr!Nleftclosedrverbose random_state) fit_interceptmax_itertolshufflerarb warm_startaverage_parameter_constraintsr7-C6??333333?TMbP?r:r1?Fr.)penaltyalphaCl1_ratiorcrdrerfraepsilonrb learning_rateeta0power_tearly_stoppingvalidation_fractionn_iter_no_changergrhc||_||_| |_| |_||_||_||_||_| |_| |_ | |_ ||_ ||_ ||_ ||_||_||_||_||_||_dSrU)lossrqrvrurrrsrtrcrfrbrarwrxryrzr{rgrhrdre)rDr}rqrrrsrtrcrdrerfrarurbrvrwrxryrzr{rgrhs rFrGzBaseSGD.__init__]s0  *    * (   ,#6 0$   rHcdS)z Fit model.NrZrDXys rFfitz BaseSGD.fitsrHc.|jr|rtd|jdvr|jdkrtd|jdkr|jdkrtd||j||jdS) zValidate input params.z/early_stopping should be False with partial_fit)r0r2r3rozeta0 must be > 0r1rzgalpha must be > 0 since learning_rate is 'optimal'. alpha is used to compute the optimal learning rate.N)ry ValueErrorrvrwrr_get_penalty_typerq_get_learning_rate_type)rDfor_partial_fits rF_more_validate_paramszBaseSGD._more_validate_paramss   P? PNOO O  "H H H S  /00 0   * *tzQ8  t|,,, $$T%788888rHcb|j|}|d|dd}}|dvr|jf}||S)z6Get concrete ``LossFunction`` object for str ``loss``.rr!N)huberepsilon_insensitivesquared_epsilon_insensitive)loss_functionsru)rDr}loss_ loss_classargss rF_get_loss_functionzBaseSGD._get_loss_functionsF#D) 8U122YD R R RL?Dz4  rHct|SrU)LEARNING_RATE_TYPES)rDrvs rFrzBaseSGD._get_learning_rate_types "=11rHc^t|}t|SrU)strlower PENALTY_TYPES)rDrqs rFrzBaseSGD._get_penalty_types$g,,$$&&W%%rHc|dkr|;tj||d}|j||fkrtd||_ntj||f|d|_|;tj|d|}|j|fkrtd||_n:tj||d|_n|Ntj||d}|}|j|fkrtd||_ntj||d|_|stj|| }|jd kr|jd krtd|r|d |_ nV|d |_n;|rtjd |d|_ ntjd |d|_|j d krx|j|_ tj|jj|d|_ |rd |j z |_ n |j|_ tj|j j|d|_dSdS)z4Allocate mem for parameters; initialize if provided.rNrsdtypeorderz+Provided ``coef_`` does not match dataset. )rrz/Provided intercept_init does not match dataset.z*Provided coef_init does not match dataset.r)r!rZr!r)rMasarrayshaperrLzerosrOravelrKoffset_rh_standard_coef _average_coef_standard_intercept_average_intercept)rD n_classes n_features input_dtype coef_initintercept_init one_classs rF_allocate_parameter_memzBaseSGD._allocate_parameter_mems q==$Jy 3OOO ?y*&===$%RSSS& X +;c )!#"#[""""'I<77$%VWWW"0"$(9Ks"S"S"S$Jy 3OOO %OO-- ?zm33$%QRRR& Xj 3OOO )!#N+!N!N!N!'4//N4HB4N4N$%VWWW#1#9#9$$DLL'5&<&<''DOOP#%8A[#L#L#LDLL&(hq 3&O&O&ODO >> J $ $D      -+++++Z^9999(!!!222&&&KKKKZ555p=A       rHr\) metaclassTcxtj|j|d}|rd|||j|k<nd|||j|k<d}d}t |jdkr{|js'|j}|jd} n|j }|j d} |j }|j d}nV|js|j|}|j|} n4|j |}|j |} |j |}|j |}||| ||fS)zeInitialization for fit_binary. Returns y, coef, intercept, average_coef, average_intercept. rsrrogrNr) rMonesrr@lenrhrLrrOrrrr) rSrir label_encodey_iaverage_intercept average_coefrQrRs rF_prepare_fit_binaryrDsL '!'C 8 8 8C)$'Aa !!%)Aa !L 3<A{ :9??$$Dq)II%++--D/2I,2244L # 6q 9  { :9Q|tjkrtntSrU)rMfloat32r*r+rs rFrrs&"*44<<,FrHceZdZUedfedfedfefefefee fe e fe e fd Z ie jeee gdgeedddgeedd d ged ged hed gd Zeed <e d$dddddddde d d ddddddd dddfd ZdZ d%dZdZdZed d&d!Zed d%d"Zfd#ZxZ S)'BaseSGDClassifierrkro) hinge squared_hinge perceptronlog_lossmodified_huber squared_errorrrrr]rr!neitherr_Nr^balanced)r}ryrzr{n_jobs class_weightrirr7rjrlTrmrnr1rpFr:r.rqrrrtrcrdrerfrarurrbrvrwrxryrzr{rrgrhct||||||||| | | | |||||||||_| |_dSN)r}rqrrrtrcrdrerfrarurbrvrwrxryrzr{rgrh)superrGrrrDr}rqrrrtrcrdrerfrarurrbrvrwrxryrzr{rrgrh __class__s rFrGzBaseSGDClassifier.__init__ st4 '%') 3-!'    *) rHc t|d } t|||dtjtjgdd| \}}| rKt |jttjfs%|jdkrtj dt|j \} }t|||jj d}t|j|j||_t%| ||j } t)|d d| ||||j| | n:||jj d kr$t/d ||jj d fz|||_t|dsd|_|dkr||||||| |n5|dkr||||||| |nt/d|z|S)Nr@csrrsF accept_sparserraccept_large_sparseresetrrPassing average=0 to disable averaging is deprecated and will be removed in 1.7. Please use average=False instead.)rErrrLrrrrrrJ6Number of features %d does not match previous data %d.r?rkrrrrsrvrrdz>The number of classes has to be greater than one; got %d class)hasattrr rMfloat64rrrhboolrwarningswarn FutureWarningrrr@rr_expanded_class_weightrrgetattrrrLrrrr?_fit_multiclass _fit_binary)rDrrrrrsr}rvrdrErrr first_callrrrs rF _partial_fitzBaseSGDClassifier._partial_fit>sr!z222   :rz* %    1  dlT28,<== $,RSBSBS P" !" :%dG444M'* ';  t}' ' ' #-]AQWMMM 4$ ' ' /93H  ( (#%G#- )    4:+B/ / /Htz/345  $66t<<tT"" DG q==  ++! !    !^^   ++!     P   rHc t|drt|dt|jtt jfs%|jdkrtjdtt||}t j |} |j r#t|dr||j }||j}nd|_ d|_|jdkr&|j |_|j|_d|_d|_d|_||||||||j| | || |j@|jt j kr*|j|jkrtjdt2|S)Nr@rr)rrLrkhMaximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.)rdelattrrrhrrMrrrrr uniquergrLrOrrrrr?r rdrerrr) rDrrrrrsr}rvrrrrEs rF_fitzBaseSGDClassifier._fits 4 $ $ & D* % % %$,rx(899 dla>O>O MH     $! $ $ $)A,, ? #wtW55 #  J %!%DJ"DO > "&"9!088B?? +-=+C+C("&":a,,DJ mI66DOOOrHc dktj}|tt j} tjj df dt| D} d} t| D]#\} \} }}|j | <t| |} $xj | jdzz c_ | _jdkrejj dz krj_j_ d Sj_t+jj _j_ d Sd S) zFit a multi-class classifier by combining binary classifiers Each binary classifier predicts one class versus all others. This strategy is called OvA (One versus All) or OvR (One versus Rest). rr)size sharedmem)rrarequirec3 K|]<\}}tt |  j|d | V=dS)rk)rrbN)rrr) .0rrrsrrrrvrdrrDrrs rF z4BaseSGDClassifier._fit_multiclass.. s  4 GJ  +A. /!         rHrorkN)rrrbrrrr@rrra enumeraterOmaxr?rrrhrrLrrrMrNr)rDrrrrrsrvrrdrbseedsresultrr_rRn_iter_irs```````` @rFrz!BaseSGDClassifier._fit_multiclasss55a]UVEV5WW *$*;<< $$W3t}3E3E$FF ; k               %U++!     ,+4V+<+< - - 'A'9h!*DOA '8,,GG 7QWQZ'' $$$D/  ///////^/b^^^PLLLL\ 7 7 7D5;5;5;n\555= = = 65= ~\555( ( ( 65( TrHrceZdZUdZiejehddgeedddgeedddgeeddd geedddgehd e ed d hgeedddgd Ze e d< d!dddddddde ddddddddddddfd Z dZeedZeed ZxZS)" SGDClassifiera.Linear classifiers (SVM, logistic regression, etc.) with SGD training. This estimator implements regularized linear models with stochastic gradient descent (SGD) learning: the gradient of the loss is estimated each sample at a time and the model is updated along the way with a decreasing strength schedule (aka learning rate). SGD allows minibatch (online/out-of-core) learning via the `partial_fit` method. For best results using the default learning rate schedule, the data should have zero mean and unit variance. This implementation works with data represented as dense or sparse arrays of floating point values for the features. The model it fits can be controlled with the loss parameter; by default, it fits a linear support vector machine (SVM). The regularizer is a penalty added to the loss function that shrinks model parameters towards the zero vector using either the squared euclidean norm L2 or the absolute norm L1 or a combination of both (Elastic Net). If the parameter update crosses the 0.0 value because of the regularizer, the update is truncated to 0.0 to allow for learning sparse models and achieve online feature selection. Read more in the :ref:`User Guide `. Parameters ---------- loss : {'hinge', 'log_loss', 'modified_huber', 'squared_hinge', 'perceptron', 'squared_error', 'huber', 'epsilon_insensitive', 'squared_epsilon_insensitive'}, default='hinge' The loss function to be used. - 'hinge' gives a linear SVM. - 'log_loss' gives logistic regression, a probabilistic classifier. - 'modified_huber' is another smooth loss that brings tolerance to outliers as well as probability estimates. - 'squared_hinge' is like hinge but is quadratically penalized. - 'perceptron' is the linear loss used by the perceptron algorithm. - The other losses, 'squared_error', 'huber', 'epsilon_insensitive' and 'squared_epsilon_insensitive' are designed for regression but can be useful in classification as well; see :class:`~sklearn.linear_model.SGDRegressor` for a description. More details about the losses formulas can be found in the :ref:`User Guide ` and you can find a visualisation of the loss functions in :ref:`sphx_glr_auto_examples_linear_model_plot_sgd_loss_functions.py`. penalty : {'l2', 'l1', 'elasticnet', None}, default='l2' The penalty (aka regularization term) to be used. Defaults to 'l2' which is the standard regularizer for linear SVM models. 'l1' and 'elasticnet' might bring sparsity to the model (feature selection) not achievable with 'l2'. No penalty is added when set to `None`. You can see a visualisation of the penalties in :ref:`sphx_glr_auto_examples_linear_model_plot_sgd_penalties.py`. alpha : float, default=0.0001 Constant that multiplies the regularization term. The higher the value, the stronger the regularization. Also used to compute the learning rate when `learning_rate` is set to 'optimal'. Values must be in the range `[0.0, inf)`. l1_ratio : float, default=0.15 The Elastic Net mixing parameter, with 0 <= l1_ratio <= 1. l1_ratio=0 corresponds to L2 penalty, l1_ratio=1 to L1. Only used if `penalty` is 'elasticnet'. Values must be in the range `[0.0, 1.0]`. fit_intercept : bool, default=True Whether the intercept should be estimated or not. If False, the data is assumed to be already centered. max_iter : int, default=1000 The maximum number of passes over the training data (aka epochs). It only impacts the behavior in the ``fit`` method, and not the :meth:`partial_fit` method. Values must be in the range `[1, inf)`. .. versionadded:: 0.19 tol : float or None, default=1e-3 The stopping criterion. If it is not None, training will stop when (loss > best_loss - tol) for ``n_iter_no_change`` consecutive epochs. Convergence is checked against the training loss or the validation loss depending on the `early_stopping` parameter. Values must be in the range `[0.0, inf)`. .. versionadded:: 0.19 shuffle : bool, default=True Whether or not the training data should be shuffled after each epoch. verbose : int, default=0 The verbosity level. Values must be in the range `[0, inf)`. epsilon : float, default=0.1 Epsilon in the epsilon-insensitive loss functions; only if `loss` is 'huber', 'epsilon_insensitive', or 'squared_epsilon_insensitive'. For 'huber', determines the threshold at which it becomes less important to get the prediction exactly right. For epsilon-insensitive, any differences between the current prediction and the correct label are ignored if they are less than this threshold. Values must be in the range `[0.0, inf)`. n_jobs : int, default=None The number of CPUs to use to do the OVA (One Versus All, for multi-class problems) computation. ``None`` means 1 unless in a :obj:`joblib.parallel_backend` context. ``-1`` means using all processors. See :term:`Glossary ` for more details. random_state : int, RandomState instance, default=None Used for shuffling the data, when ``shuffle`` is set to ``True``. Pass an int for reproducible output across multiple function calls. See :term:`Glossary `. Integer values must be in the range `[0, 2**32 - 1]`. learning_rate : str, default='optimal' The learning rate schedule: - 'constant': `eta = eta0` - 'optimal': `eta = 1.0 / (alpha * (t + t0))` where `t0` is chosen by a heuristic proposed by Leon Bottou. - 'invscaling': `eta = eta0 / pow(t, power_t)` - 'adaptive': `eta = eta0`, as long as the training keeps decreasing. Each time n_iter_no_change consecutive epochs fail to decrease the training loss by tol or fail to increase validation score by tol if `early_stopping` is `True`, the current learning rate is divided by 5. .. versionadded:: 0.20 Added 'adaptive' option. eta0 : float, default=0.0 The initial learning rate for the 'constant', 'invscaling' or 'adaptive' schedules. The default value is 0.0 as eta0 is not used by the default schedule 'optimal'. Values must be in the range `[0.0, inf)`. power_t : float, default=0.5 The exponent for inverse scaling learning rate. Values must be in the range `(-inf, inf)`. early_stopping : bool, default=False Whether to use early stopping to terminate training when validation score is not improving. If set to `True`, it will automatically set aside a stratified fraction of training data as validation and terminate training when validation score returned by the `score` method is not improving by at least tol for n_iter_no_change consecutive epochs. See :ref:`sphx_glr_auto_examples_linear_model_plot_sgd_early_stopping.py` for an example of the effects of early stopping. .. versionadded:: 0.20 Added 'early_stopping' option validation_fraction : float, default=0.1 The proportion of training data to set aside as validation set for early stopping. Must be between 0 and 1. Only used if `early_stopping` is True. Values must be in the range `(0.0, 1.0)`. .. versionadded:: 0.20 Added 'validation_fraction' option n_iter_no_change : int, default=5 Number of iterations with no improvement to wait before stopping fitting. Convergence is checked against the training loss or the validation loss depending on the `early_stopping` parameter. Integer values must be in the range `[1, max_iter)`. .. versionadded:: 0.20 Added 'n_iter_no_change' option class_weight : dict, {class_label: weight} or "balanced", default=None Preset for the class_weight fit parameter. Weights associated with classes. If not given, all classes are supposed to have weight one. 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))``. warm_start : bool, default=False When set to True, reuse the solution of the previous call to fit as initialization, otherwise, just erase the previous solution. See :term:`the Glossary `. Repeatedly calling fit or partial_fit when warm_start is True can result in a different solution than when calling fit a single time because of the way the data is shuffled. If a dynamic learning rate is used, the learning rate is adapted depending on the number of samples already seen. Calling ``fit`` resets this counter, while ``partial_fit`` will result in increasing the existing counter. average : bool or int, default=False When set to `True`, computes the averaged SGD weights across all updates and stores the result in the ``coef_`` attribute. If set to an int greater than 1, averaging will begin once the total number of samples seen reaches `average`. So ``average=10`` will begin averaging after seeing 10 samples. Integer values must be in the range `[1, n_samples]`. Attributes ---------- coef_ : ndarray of shape (1, n_features) if n_classes == 2 else (n_classes, n_features) Weights assigned to the features. intercept_ : ndarray of shape (1,) if n_classes == 2 else (n_classes,) Constants in decision function. n_iter_ : int The actual number of iterations before reaching the stopping criterion. For multiclass fits, it is the maximum over every binary fit. classes_ : array of shape (n_classes,) t_ : int Number of weight updates performed during training. Same as ``(n_iter_ * n_samples + 1)``. n_features_in_ : int Number of features seen during :term:`fit`. .. versionadded:: 0.24 feature_names_in_ : ndarray of shape (`n_features_in_`,) Names of features seen during :term:`fit`. Defined only when `X` has feature names that are all strings. .. versionadded:: 1.0 See Also -------- sklearn.svm.LinearSVC : Linear support vector classification. LogisticRegression : Logistic regression. Perceptron : Inherits from SGDClassifier. ``Perceptron()`` is equivalent to ``SGDClassifier(loss="perceptron", eta0=1, learning_rate="constant", penalty=None)``. Examples -------- >>> import numpy as np >>> from sklearn.linear_model import SGDClassifier >>> from sklearn.preprocessing import StandardScaler >>> from sklearn.pipeline import make_pipeline >>> X = np.array([[-1, -1], [-2, -1], [1, 1], [2, 1]]) >>> Y = np.array([1, 1, 2, 2]) >>> # Always scale the input. The most convenient way is to use a pipeline. >>> clf = make_pipeline(StandardScaler(), ... SGDClassifier(max_iter=1000, tol=1e-3)) >>> clf.fit(X, Y) Pipeline(steps=[('standardscaler', StandardScaler()), ('sgdclassifier', SGDClassifier())]) >>> print(clf.predict([[-0.8, -1]])) [1] >r8r7r9Nrr^r_r!bothr>r1r3r0r2r4r5)rqrrrtrxrurvrwrirr7rjrlTrmrnr1rorpFr:r.rctt||||||||| | | | | ||||||||dS)N)r}rqrrrtrcrdrerfrarurrbrvrwrxryrzr{rrgrhrrGrs rFrGzSGDClassifier.__init__so2 '%') 3-%!+      rHcF|jdvrtd|jzdS)N)rrz3probability estimates are not available for loss=%rT)r}AttributeError)rDs rF _check_probazSGDClassifier._check_probas3 9: : : E Q trHct||jdkr||S|jdkrt|jdk}||}|r.t j|jddf}|dddf}n|}t j |dd||dz }|d z}|r|dddfxx|zcc<|}nu| d }|dk}t j |r d||ddf<t|j||<|| |jddfz}|Std |jz) ahProbability estimates. This method is only available for log loss and modified Huber loss. Multiclass probability estimates are derived from binary (one-vs.-rest) estimates by simple normalization, as recommended by Zadrozny and Elkan. Binary probability estimates for loss="modified_huber" are given by (clip(decision_function(X), -1, 1) + 1) / 2. For other loss functions it is necessary to perform proper probability calibration by wrapping the classifier with :class:`~sklearn.calibration.CalibratedClassifierCV` instead. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Input data for prediction. Returns ------- ndarray of shape (n_samples, n_classes) Returns the probability of the sample for each class in the model, where classes are ordered as they are in `self.classes_`. References ---------- Zadrozny and Elkan, "Transforming classifier scores into multiclass probability estimates", SIGKDD'02, https://dl.acm.org/doi/pdf/10.1145/775047.775151 The justification for the formula in the loss="modified_huber" case is in the appendix B in: http://jmlr.csail.mit.edu/papers/volume2/zhang02c/zhang02c.pdf rrrrNr!rJrkg@)axisz[predict_(log_)proba only supported when loss='log_loss' or loss='modified_huber' (%r given))rr}_predict_proba_lrrr@decision_functionrMrrclipsumrrKNotImplementedError)rDrbinaryscoresprob2probprob_sumall_zeros rF predict_probazSGDClassifier.predict_probasJ  9 " "))!,, , Y* * *''1,F++A..F a! 455QQQT{ GFB4 ( ( ( CKD CKD >aaad t#   888++#q=6(##<()D111%),T]););HX&(($*Q-)<===K&#y) rHcPtj||S)aLog of probability estimates. This method is only available for log loss and modified Huber loss. When loss="modified_huber", probability estimates may be hard zeros and ones, so taking the logarithm is not possible. See ``predict_proba`` for details. Parameters ---------- X : {array-like, sparse matrix} of shape (n_samples, n_features) Input data for prediction. Returns ------- T : array-like, shape (n_samples, n_classes) Returns the log-probability of the sample for each class in the model, where classes are ordered as they are in `self.classes_`. )rMlogrNrDrs rFpredict_log_probazSGDClassifier.predict_log_probaUs".vd((++,,,rHr2)rVrWrXrYrrirrrrrrr5rGr@rrNrRr7r8s@rFr:r:sEEN $  2 $J999::DA(4D8889XdAq8889HT4i@@@AHT1d6:::; JHHH I I F::uen-- . . $47778 $ $ $D    /   // / / / / / / b\,NN N`\,-- -----rHr:ceZdZUefeefeefeefdZie j e e egdge edddge edddgd Z eed <e d#d d ddddddeddddddddddfd ZdZedd$dZ d%dZedd%dZdZd Zd!Zfd"ZxZS)&BaseSGDRegressor)rrrrr]rr!rr_Nr^)r}ryrzr{rirr7rjrlTrmrnr2{Gz??Fr:r.rqrrrtrcrdrerfrarurbrvrwrxryrzr{rgrhcpt||||||||| | | | | ||||||dSrr=rDr}rqrrrtrcrdrerfrarurbrvrwrxryrzr{rgrhrs rFrGzBaseSGDRegressor.__init__si0 '%') 3-!'      rHc t|dddu} t|||dddtjtjgd|  \}}||jd}| rKt|jttj fs%|jdkrtj dt|j\} } t|||j }| r|d | |j| | |jdkrSt|d dBtj| |jd |_tjd |jd |_||||||||||S)NrLrFrs)rcopyrrrr)r[rrrr!rrr)rr rMrrastyperrrhrrrrrrrrrrr_fit_regressor)rDrrrrrsr}rvrdrrrr rrs rFr zBaseSGDRegressor._partial_fitsT7D11T9   :rz* %    1 HHQW5H ) )  dlT28,<== $,RSBSBS P" !" :,]AQWMMM     ( (%G#- )    O>O MH     ? #wtWd;;G  J %!%DJ"DO     M     H BF7""  -- M'#     rHc |||||jd|j|j||| S)aFit linear model with Stochastic Gradient Descent. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Training data. y : ndarray of shape (n_samples,) Target values. coef_init : ndarray of shape (n_features,), default=None The initial coefficients to warm-start the optimization. intercept_init : ndarray of shape (1,), default=None The initial intercept to warm-start the optimization. sample_weight : array-like, shape (n_samples,), default=None Weights applied to individual samples (1. for unweighted). Returns ------- self : object Fitted `SGDRegressor` estimator. rkr&r'r(s rFrzBaseSGDRegressor.fitSsS4 ""$$$yy *,)'   rHct|t||dd}t||jjd|jz}|S)aPredict using the linear model Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Returns ------- ndarray of shape (n_samples,) Predicted target values per element in X. rFrrT dense_output)rr rrLTrOr)rDrrIs rF_decision_functionz#BaseSGDRegressor._decision_function{sU  $e D D D DJLtDDDtV||~~rHc,||S)a*Predict using the linear model. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Input data. Returns ------- ndarray of shape (n_samples,) Predicted target values per element in X. )rgrQs rFpredictzBaseSGDRegressor.predicts&&q)))rHc b||} ||j} ||} t |dsd|_|||dk} || |||} t|j }| dt}t||||\}}|j |j n tj }|jr|j}|j}|j}|j}n|j}|j}d}dg}t/|j}|||d||d| | |||j|| |j| t7|j||t7|jt7|jt7|j|dd| |j |j!d|j||j\}}}}|_"|xj|j"|j#dzz c_|jdkrtj$||_tj$||_|j|jdz kr"||_tj$||_dS||_tj$||_dStj$||_dS)Nr?rkrrrr)%rrrqrrr?rrrrbrrr$rerMrrhrrrrrLrOrrrtryrr{rcrarfrwrxrrrN)rDrrrrrsr}rvrrd loss_functionrrrrrbrrrrerQrRrrrs rFr]zBaseSGDRegressor._fit_regressors//55 --dl;; !99-HHtT"" DG55a]UVEV5WW"<< Q=  *$*;<< ##Aw//#/ q-l$ $ $  (.dhhRVG < $&D0I-L $ 7  :DIL!" ,DDD IS  aL  a    M      % & &   " # #           I L G  L;J J Fi'8$,@ 4`. Parameters ---------- loss : str, default='squared_error' The loss function to be used. The possible values are 'squared_error', 'huber', 'epsilon_insensitive', or 'squared_epsilon_insensitive' The 'squared_error' refers to the ordinary least squares fit. 'huber' modifies 'squared_error' to focus less on getting outliers correct by switching from squared to linear loss past a distance of epsilon. 'epsilon_insensitive' ignores errors less than epsilon and is linear past that; this is the loss function used in SVR. 'squared_epsilon_insensitive' is the same but becomes squared loss past a tolerance of epsilon. More details about the losses formulas can be found in the :ref:`User Guide `. penalty : {'l2', 'l1', 'elasticnet', None}, default='l2' The penalty (aka regularization term) to be used. Defaults to 'l2' which is the standard regularizer for linear SVM models. 'l1' and 'elasticnet' might bring sparsity to the model (feature selection) not achievable with 'l2'. No penalty is added when set to `None`. You can see a visualisation of the penalties in :ref:`sphx_glr_auto_examples_linear_model_plot_sgd_penalties.py`. alpha : float, default=0.0001 Constant that multiplies the regularization term. The higher the value, the stronger the regularization. Also used to compute the learning rate when `learning_rate` is set to 'optimal'. Values must be in the range `[0.0, inf)`. l1_ratio : float, default=0.15 The Elastic Net mixing parameter, with 0 <= l1_ratio <= 1. l1_ratio=0 corresponds to L2 penalty, l1_ratio=1 to L1. Only used if `penalty` is 'elasticnet'. Values must be in the range `[0.0, 1.0]`. fit_intercept : bool, default=True Whether the intercept should be estimated or not. If False, the data is assumed to be already centered. max_iter : int, default=1000 The maximum number of passes over the training data (aka epochs). It only impacts the behavior in the ``fit`` method, and not the :meth:`partial_fit` method. Values must be in the range `[1, inf)`. .. versionadded:: 0.19 tol : float or None, default=1e-3 The stopping criterion. If it is not None, training will stop when (loss > best_loss - tol) for ``n_iter_no_change`` consecutive epochs. Convergence is checked against the training loss or the validation loss depending on the `early_stopping` parameter. Values must be in the range `[0.0, inf)`. .. versionadded:: 0.19 shuffle : bool, default=True Whether or not the training data should be shuffled after each epoch. verbose : int, default=0 The verbosity level. Values must be in the range `[0, inf)`. epsilon : float, default=0.1 Epsilon in the epsilon-insensitive loss functions; only if `loss` is 'huber', 'epsilon_insensitive', or 'squared_epsilon_insensitive'. For 'huber', determines the threshold at which it becomes less important to get the prediction exactly right. For epsilon-insensitive, any differences between the current prediction and the correct label are ignored if they are less than this threshold. Values must be in the range `[0.0, inf)`. random_state : int, RandomState instance, default=None Used for shuffling the data, when ``shuffle`` is set to ``True``. Pass an int for reproducible output across multiple function calls. See :term:`Glossary `. learning_rate : str, default='invscaling' The learning rate schedule: - 'constant': `eta = eta0` - 'optimal': `eta = 1.0 / (alpha * (t + t0))` where t0 is chosen by a heuristic proposed by Leon Bottou. - 'invscaling': `eta = eta0 / pow(t, power_t)` - 'adaptive': eta = eta0, as long as the training keeps decreasing. Each time n_iter_no_change consecutive epochs fail to decrease the training loss by tol or fail to increase validation score by tol if early_stopping is True, the current learning rate is divided by 5. .. versionadded:: 0.20 Added 'adaptive' option. eta0 : float, default=0.01 The initial learning rate for the 'constant', 'invscaling' or 'adaptive' schedules. The default value is 0.01. Values must be in the range `[0.0, inf)`. power_t : float, default=0.25 The exponent for inverse scaling learning rate. Values must be in the range `(-inf, inf)`. early_stopping : bool, default=False Whether to use early stopping to terminate training when validation score is not improving. If set to True, it will automatically set aside a fraction of training data as validation and terminate training when validation score returned by the `score` method is not improving by at least `tol` for `n_iter_no_change` consecutive epochs. See :ref:`sphx_glr_auto_examples_linear_model_plot_sgd_early_stopping.py` for an example of the effects of early stopping. .. versionadded:: 0.20 Added 'early_stopping' option validation_fraction : float, default=0.1 The proportion of training data to set aside as validation set for early stopping. Must be between 0 and 1. Only used if `early_stopping` is True. Values must be in the range `(0.0, 1.0)`. .. versionadded:: 0.20 Added 'validation_fraction' option n_iter_no_change : int, default=5 Number of iterations with no improvement to wait before stopping fitting. Convergence is checked against the training loss or the validation loss depending on the `early_stopping` parameter. Integer values must be in the range `[1, max_iter)`. .. versionadded:: 0.20 Added 'n_iter_no_change' option warm_start : bool, default=False When set to True, reuse the solution of the previous call to fit as initialization, otherwise, just erase the previous solution. See :term:`the Glossary `. Repeatedly calling fit or partial_fit when warm_start is True can result in a different solution than when calling fit a single time because of the way the data is shuffled. If a dynamic learning rate is used, the learning rate is adapted depending on the number of samples already seen. Calling ``fit`` resets this counter, while ``partial_fit`` will result in increasing the existing counter. average : bool or int, default=False When set to True, computes the averaged SGD weights across all updates and stores the result in the ``coef_`` attribute. If set to an int greater than 1, averaging will begin once the total number of samples seen reaches `average`. So ``average=10`` will begin averaging after seeing 10 samples. Attributes ---------- coef_ : ndarray of shape (n_features,) Weights assigned to the features. intercept_ : ndarray of shape (1,) The intercept term. n_iter_ : int The actual number of iterations before reaching the stopping criterion. t_ : int Number of weight updates performed during training. Same as ``(n_iter_ * n_samples + 1)``. n_features_in_ : int Number of features seen during :term:`fit`. .. versionadded:: 0.24 feature_names_in_ : ndarray of shape (`n_features_in_`,) Names of features seen during :term:`fit`. Defined only when `X` has feature names that are all strings. .. versionadded:: 1.0 See Also -------- HuberRegressor : Linear regression model that is robust to outliers. Lars : Least Angle Regression model. Lasso : Linear Model trained with L1 prior as regularizer. RANSACRegressor : RANSAC (RANdom SAmple Consensus) algorithm. Ridge : Linear least squares with l2 regularization. sklearn.svm.SVR : Epsilon-Support Vector Regression. TheilSenRegressor : Theil-Sen Estimator robust multivariate regression model. Examples -------- >>> import numpy as np >>> from sklearn.linear_model import SGDRegressor >>> from sklearn.pipeline import make_pipeline >>> from sklearn.preprocessing import StandardScaler >>> n_samples, n_features = 10, 5 >>> rng = np.random.RandomState(0) >>> y = rng.randn(n_samples) >>> X = rng.randn(n_samples, n_features) >>> # Always scale the input. The most convenient way is to use a pipeline. >>> reg = make_pipeline(StandardScaler(), ... SGDRegressor(max_iter=1000, tol=1e-3)) >>> reg.fit(X, y) Pipeline(steps=[('standardscaler', StandardScaler()), ('sgdregressor', SGDRegressor())]) >r8r7r9Nrr^r_r!r;r>r1r3r0r2r4r5)rqrrrtrxrvrurwrirr7rjrlTrmrnr2rUrVFr:r.rWcpt||||||||| | | | | ||||||dSrr=rYs rFrGzSGDRegressor.__init__si. '%') 3-!'      rHrm)rVrWrXrYrTrirrrrrrr5rGr7r8s@rFrorosddL $  1 $J999::DA(4D8889XdAq8889HT4i@@@A JHHH I I F::uen-- . . HT1d6:::;$47778 $ $ $D    +  " ++ + + + + + + + + + + rHroc eZdZUdZdedfiZiejee dddge hde e dd hgee d d d gee d d d gdZe e d< d!fd ZdZdZedd"dZ d#dZedd$dZdZdZdZfd ZxZS)%SGDOneClassSVMaSolves linear One-Class SVM using Stochastic Gradient Descent. This implementation is meant to be used with a kernel approximation technique (e.g. `sklearn.kernel_approximation.Nystroem`) to obtain results similar to `sklearn.svm.OneClassSVM` which uses a Gaussian kernel by default. Read more in the :ref:`User Guide `. .. versionadded:: 1.0 Parameters ---------- nu : float, default=0.5 The nu parameter of the One Class SVM: an upper bound on the fraction of training errors and a lower bound of the fraction of support vectors. Should be in the interval (0, 1]. By default 0.5 will be taken. fit_intercept : bool, default=True Whether the intercept should be estimated or not. Defaults to True. max_iter : int, default=1000 The maximum number of passes over the training data (aka epochs). It only impacts the behavior in the ``fit`` method, and not the `partial_fit`. Defaults to 1000. Values must be in the range `[1, inf)`. tol : float or None, default=1e-3 The stopping criterion. If it is not None, the iterations will stop when (loss > previous_loss - tol). Defaults to 1e-3. Values must be in the range `[0.0, inf)`. shuffle : bool, default=True Whether or not the training data should be shuffled after each epoch. Defaults to True. verbose : int, default=0 The verbosity level. random_state : int, RandomState instance or None, 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`. learning_rate : {'constant', 'optimal', 'invscaling', 'adaptive'}, default='optimal' The learning rate schedule to use with `fit`. (If using `partial_fit`, learning rate must be controlled directly). - 'constant': `eta = eta0` - 'optimal': `eta = 1.0 / (alpha * (t + t0))` where t0 is chosen by a heuristic proposed by Leon Bottou. - 'invscaling': `eta = eta0 / pow(t, power_t)` - 'adaptive': eta = eta0, as long as the training keeps decreasing. Each time n_iter_no_change consecutive epochs fail to decrease the training loss by tol or fail to increase validation score by tol if early_stopping is True, the current learning rate is divided by 5. eta0 : float, default=0.0 The initial learning rate for the 'constant', 'invscaling' or 'adaptive' schedules. The default value is 0.0 as eta0 is not used by the default schedule 'optimal'. Values must be in the range `[0.0, inf)`. power_t : float, default=0.5 The exponent for inverse scaling learning rate. Values must be in the range `(-inf, inf)`. warm_start : bool, default=False When set to True, reuse the solution of the previous call to fit as initialization, otherwise, just erase the previous solution. See :term:`the Glossary `. Repeatedly calling fit or partial_fit when warm_start is True can result in a different solution than when calling fit a single time because of the way the data is shuffled. If a dynamic learning rate is used, the learning rate is adapted depending on the number of samples already seen. Calling ``fit`` resets this counter, while ``partial_fit`` will result in increasing the existing counter. average : bool or int, default=False When set to True, computes the averaged SGD weights and stores the result in the ``coef_`` attribute. If set to an int greater than 1, averaging will begin once the total number of samples seen reaches average. So ``average=10`` will begin averaging after seeing 10 samples. Attributes ---------- coef_ : ndarray of shape (1, n_features) Weights assigned to the features. offset_ : ndarray of shape (1,) Offset used to define the decision function from the raw scores. We have the relation: decision_function = score_samples - offset. n_iter_ : int The actual number of iterations to reach the stopping criterion. t_ : int Number of weight updates performed during training. Same as ``(n_iter_ * n_samples + 1)``. n_features_in_ : int Number of features seen during :term:`fit`. .. versionadded:: 0.24 feature_names_in_ : ndarray of shape (`n_features_in_`,) Names of features seen during :term:`fit`. Defined only when `X` has feature names that are all strings. .. versionadded:: 1.0 See Also -------- sklearn.svm.OneClassSVM : Unsupervised Outlier Detection. Notes ----- This estimator has a linear complexity in the number of training samples and is thus better suited than the `sklearn.svm.OneClassSVM` implementation for datasets with a large number of training samples (say > 10,000). Examples -------- >>> import numpy as np >>> from sklearn import linear_model >>> X = np.array([[-1, -1], [-2, -1], [1, 1], [2, 1]]) >>> clf = linear_model.SGDOneClassSVM(random_state=42) >>> clf.fit(X) SGDOneClassSVM(random_state=42) >>> print(clf.predict([[4, 4]])) [1] rrkrorightr_>r1r3r0r2r4r5rNr^r)nurvrwrxrirpTrmrnr1Fc ||_tt|dddd|||||t||| | ddd| | dS) Nrr7rkrFr:r.)r}rqrsrtrcrdrerfrarurbrvrwrxryrzr{rgrh)rtrrrrGr5)rDrtrcrdrerfrarbrvrwrxrgrhrs rFrGzSGDOneClassSVM.__init__sv nd##,,'#%' #!' -     rHc|jd}tj||jd}t |||\} } ||j} ||} |||dk} | | |||}t|j }| dtj tjj}|j|jn tj }d}d}d}|jr|j}|j}|j}|j}n|j}d|jz }d}dg}t3|j}|||d||d|j| |||j| | |j|t;|j||t;|jt;|j t;|j!|||| |j"|j#||j$| |j\}}}}|_%|xj$|j%|zz c_$|jdkrtj&||_tj&||_|j|j$dz kr%||_dtj&|z |_dS||_dtj&|z |_dSdtj&|z |_dS) z8Uses SGD implementation with X and y=np.ones(n_samples).rrsrrNr!rrk)'rrMrrr$rrqrrrrrbriinfoint32rrerrhrrrrrLrrrrtryrr{rcrarfrwrxr?rrN)rDrrrrsrrvrdrrr offset_decayrrrrrbrrerrrrQrRrrrs rF_fit_one_classzSGDOneClassSVM._fit_one_classs GAJ GIQWC 8 8 8 ,Q= A A--dl;; !99-HH 55a]UVEV5WW"<< Q=  *$*;<< ##Arx'9'9'=>>(.dhhRVG   < $&D0I-L $ 7  :DDL(IL!" ,DDD IS  aL  a    M      % & &   " # #           I L  G  L;J J Fi'8$,@ 4<)++  sT7D11T9   :rz* %   dlT28,<== $,RSBSBS P" WQZ -]AQWMMM 4$ ' ' /93H  ( (%G#* )    4:+B/ / /Htz/345  < LGD/4@@H!#*AG3!O!O!OD &(hqs&K&K&KD ##66t<<tT"" DG  ''     rHrc t|ds|d|jdz }|||d|j|jd|dd S) aDFit linear One-Class SVM with Stochastic Gradient Descent. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Subset of the training data. y : Ignored Not used, present for API consistency by convention. sample_weight : array-like, shape (n_samples,), optional Weights applied to individual samples. If not provided, uniform weights are assumed. Returns ------- self : object Returns a fitted instance of self. rLTr"rrkr!N)rsr}rvrdrrr|)rrrtr r}rv)rDrrrrrs rFr$zSGDOneClassSVM.partial_fit su(tW%% =  & &t & < < <!   ,'!   rHc t|jttjfs%|jdkrt jdt|jr#t|dr||j }||j }nd|_ d|_ d|_ | ||||||j||| |j@|jtj kr*|j|jkrt jdt$|Sr`)rrhrrMrrrrrgrrLrr?r rdrerrr) rDrrrrsr}rvrr|rs rFrzSGDOneClassSVM._fit s$,rx(899 dla>O>O MH     ? wtW55   J ""l DJDL     M     H BF7""  -- M'#     rHc ||jdz }|||d|j|j||||S)aFit linear One-Class SVM with Stochastic Gradient Descent. This solves an equivalent optimization problem of the One-Class SVM primal optimization problem and returns a weight vector w and an offset rho such that the decision function is given by - rho. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Training data. y : Ignored Not used, present for API consistency by convention. coef_init : array, shape (n_classes, n_features) The initial coefficients to warm-start the optimization. offset_init : array, shape (n_classes,) The initial offset to warm-start the optimization. sample_weight : array-like, shape (n_samples,), optional Weights applied to individual samples. If not provided, uniform weights are assumed. These weights will be multiplied with class_weight (passed through the constructor) if class_weight is specified. Returns ------- self : object Returns a fitted instance of self. rrk)rrrsr}rvrr|r)rrtrr}rv)rDrrrr|rrrs rFrzSGDOneClassSVM.fit saB ""$$$!  ,#'   rHct|dt||dd}t||jjd|jz }|S)aSigned distance to the separating hyperplane. Signed distance is positive for an inlier and negative for an outlier. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Testing data. Returns ------- dec : array-like, shape (n_samples,) Decision function values of the samples. rLrFrcTrd)rr rrLrfrr)rDr decisionss rFrDz SGDOneClassSVM.decision_function sY" g&&& $e D D D#Atz|$GGG$,V    rHc@|||jz}|S)aLRaw scoring function of the samples. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Testing data. Returns ------- score_samples : array-like, shape (n_samples,) Unshiffted scoring function values of the samples. )rDr)rDr score_sampless rFrzSGDOneClassSVM.score_samples6 s$..q11DL@ rHc||dktj}d||dk<|S)a/Return labels (1 inlier, -1 outlier) of the samples. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Testing data. Returns ------- y : array, shape (n_samples,) Labels of the samples. rrJ)rDr\rMrxrs rFrizSGDOneClassSVM.predictF s> # #A & &! + 3 3BH = =!q& rHc`t}d|j_|Sr*r+r/s rFr,zSGDOneClassSVM.__sklearn_tags__W r1rH) rpTrmrnTrNr1rorpFFr4r3)NNNN)rVrWrXrYr&rr\rirrrrrrrGrzr rr$rrrDrrir,r7r8s@rFrrrrs!KKZs|,N $  ( $xc3w7778 JHHH I I F::uen-- . . $47778HT4i@@@A $ $ $D      $ $ $ $ $ $ L^8^8^8@LLL\\555! ! ! 65! T;;;;z\555...65.`!!!0 "rHrr)Tr4)KrYrabcrrnumbersrrnumpyrM _loss._lossrr r baser r r rrr exceptionsrmodel_selectionrrutilsrrutils._param_validationrrr utils.extmathrutils.metaestimatorsrutils.multiclassrutils.parallelrrutils.validationrrr _baser"r#r$ _sgd_fastr%r&r'r(r)r*r+rrr5rwrxrrr<r\rrrrr:rTrorrrZrHrFrs/''''''''""""""""MMMMMMMMMM,+++++BBBBBBBB<<<<<<<<BBBBBBBBBB++++++//////<<<<<<........SSSSSSSSSSGGGGGGGGGG !1A>>  "(28   IIIIIIII&r r r r r o}r r r r j"A"A"A"AbE$E$E$E$PGGG~~~~~-w'~~~~Bv-v-v-v-v-%v-v-v-r HHHHH~wHHHV ` ` ` ` ` #` ` ` F }}}}}\7}}}}}rH