ó ‡ˆ\c@s´ddlZddlmZmZddlmZmZddlmZddl m Z ddl m Z dd l m Z dd l mZdd „Zd eeefd „ƒYZdS(i˙˙˙˙N(toptimizetsparsei(t BaseEstimatortRegressorMixini(t LinearModel(t check_X_y(tcheck_consistent_length(taxis0_safe_slice(tsafe_sparse_dotc!Csčtj|ƒ}|j\}}|d|jdk} | rH|d} n|d} || }tj|ƒ} |t||ƒ} | r‘| | 8} ntj| ƒ}||| k}||}tj|ƒ}|jd|}||}tj|ƒ}d|tj||ƒ| ||d}| |}|||}tj|j |ƒ}|| }| rstj |dƒ}ntj |dƒ}t |||ƒ }d| t||ƒ||*tj |ƒ}| |dk}d|| epsilon``, where w and sigma are parameters to be optimized. The parameter sigma makes sure that if y is scaled up or down by a certain factor, one does not need to rescale epsilon to achieve the same robustness. Note that this does not take into account the fact that the different features of X may be of different scales. This makes sure that the loss function is not heavily influenced by the outliers while not completely ignoring their effect. Read more in the :ref:`User Guide ` .. versionadded:: 0.18 Parameters ---------- epsilon : float, greater than 1.0, default 1.35 The parameter epsilon controls the number of samples that should be classified as outliers. The smaller the epsilon, the more robust it is to outliers. max_iter : int, default 100 Maximum number of iterations that scipy.optimize.fmin_l_bfgs_b should run for. alpha : float, default 0.0001 Regularization parameter. warm_start : bool, default False This is useful if the stored attributes of a previously used model has to be reused. If set to False, then the coefficients will be rewritten for every call to fit. See :term:`the Glossary `. fit_intercept : bool, default True Whether or not to fit the intercept. This can be set to False if the data is already centered around the origin. tol : float, default 1e-5 The iteration will stop when ``max{|proj g_i | i = 1, ..., n}`` <= ``tol`` where pg_i is the i-th component of the projected gradient. Attributes ---------- coef_ : array, shape (n_features,) Features got by optimizing the Huber loss. intercept_ : float Bias. scale_ : float The value by which ``|y - X'w - c|`` is scaled down. n_iter_ : int Number of iterations that fmin_l_bfgs_b has run for. .. versionchanged:: 0.20 In SciPy <= 1.0.0 the number of lbfgs iterations may exceed ``max_iter``. ``n_iter_`` will now report at most ``max_iter``. outliers_ : array, shape (n_samples,) A boolean mask which is set to True where the samples are identified as outliers. Examples -------- >>> import numpy as np >>> from sklearn.linear_model import HuberRegressor, LinearRegression >>> from sklearn.datasets import make_regression >>> np.random.seed(0) >>> X, y, coef = make_regression( ... n_samples=200, n_features=2, noise=4.0, coef=True, random_state=0) >>> X[:4] = np.random.uniform(10, 20, (4, 2)) >>> y[:4] = np.random.uniform(10, 20, 4) >>> huber = HuberRegressor().fit(X, y) >>> huber.score(X, y) # doctest: +ELLIPSIS -7.284608623514573 >>> huber.predict(X[:1,]) array([806.7200...]) >>> linear = LinearRegression().fit(X, y) >>> print("True coefficients:", coef) True coefficients: [20.4923... 34.1698...] >>> print("Huber coefficients:", huber.coef_) Huber coefficients: [17.7906... 31.0106...] >>> print("Linear Regression coefficients:", linear.coef_) Linear Regression coefficients: [-1.9221... 7.0226...] References ---------- .. [1] Peter J. Huber, Elvezio M. Ronchetti, Robust Statistics Concomitant scale estimates, pg 172 .. [2] Art B. Owen (2006), A robust hybrid of lasso and ridge regression. http://statweb.stanford.edu/~owen/reports/hhu.pdf gš™™™™™ő?idg-Cëâ6?gńh㈾řä>cCs:||_||_||_||_||_||_dS(N(Rtmax_iterRt warm_startRttol(tselfRR6RR7RR8((s9lib/python2.7/site-packages/sklearn/linear_model/huber.pyt__init__âs      c Cs‚t||dtddgdtƒ\}}|dk rUtj|ƒ}t||ƒntj|ƒ}|jdkr‰t d|jƒ‚n|j rČt |dƒrČtj |j |j|jgfƒ}nJ|jrîtj|jdd ƒ}ntj|jddƒ}d|d s  m