1. class Lasso Found at: sklearn.linear_model._coordinate_descent
2. 
3. class Lasso(ElasticNet):
4. """Linear Model trained with L1 prior as regularizer (aka the Lasso)
5.     
6.     The optimization objective for Lasso is::
7.     
8.     (1 / (2 * n_samples)) * ||y - Xw||^2_2 + alpha * ||w||_1
9.     
10.     Technically the Lasso model is optimizing the same objective function as
11.     the Elastic Net with ``l1_ratio=1.0`` (no L2 penalty).
12.     
13.     Read more in the :ref:`User Guide <lasso>`.
14.     
15.     Parameters
16.     ----------
17.     alpha : float, default=1.0
18.     Constant that multiplies the L1 term. Defaults to 1.0.
19.     ``alpha = 0`` is equivalent to an ordinary least square, solved
20.     by the :class:`LinearRegression` object. For numerical
21.     reasons, using ``alpha = 0`` with the ``Lasso`` object is not advised.
22.     Given this, you should use the :class:`LinearRegression` object.
23.     
24.     fit_intercept : bool, default=True
25.     Whether to calculate the intercept for this model. If set
26.     to False, no intercept will be used in calculations
27.     (i.e. data is expected to be centered).
28.     
29.     normalize : bool, default=False
30.     This parameter is ignored when ``fit_intercept`` is set to False.
31.     If True, the regressors X will be normalized before regression by
32.     subtracting the mean and dividing by the l2-norm.
33.     If you wish to standardize, please use
34.     :class:`sklearn.preprocessing.StandardScaler` before calling ``fit``
35.     on an estimator with ``normalize=False``.
36.     
37.     precompute : 'auto', bool or array-like of shape (n_features, n_features),\
38.     default=False
39.     Whether to use a precomputed Gram matrix to speed up
40.     calculations. If set to ``'auto'`` let us decide. The Gram
41.     matrix can also be passed as argument. For sparse input
42.     this option is always ``True`` to preserve sparsity.
43.     
44.     copy_X : bool, default=True
45.     If ``True``, X will be copied; else, it may be overwritten.
46.     
47.     max_iter : int, default=1000
48.     The maximum number of iterations
49.     
50.     tol : float, default=1e-4
51.     The tolerance for the optimization: if the updates are
52.     smaller than ``tol``, the optimization code checks the
53.     dual gap for optimality and continues until it is smaller
54.     than ``tol``.
55.     
56.     warm_start : bool, default=False
57.     When set to True, reuse the solution of the previous call to fit as
58.     initialization, otherwise, just erase the previous solution.
59.     See :term:`the Glossary <warm_start>`.
60.     
61.     positive : bool, default=False
62.     When set to ``True``, forces the coefficients to be positive.
63.     
64.     random_state : int, RandomState instance, default=None
65.     The seed of the pseudo random number generator that selects a 
66.      random
67.     feature to update. Used when ``selection`` == 'random'.
68.     Pass an int for reproducible output across multiple function calls.
69.     See :term:`Glossary <random_state>`.
70.     
71.     selection : {'cyclic', 'random'}, default='cyclic'
72.     If set to 'random', a random coefficient is updated every iteration
73.     rather than looping over features sequentially by default. This
74.     (setting to 'random') often leads to significantly faster convergence
75.     especially when tol is higher than 1e-4.
76.     
77.     Attributes
78.     ----------
79.     coef_ : ndarray of shape (n_features,) or (n_targets, n_features)
80.     parameter vector (w in the cost function formula)
81.     
82.     sparse_coef_ : sparse matrix of shape (n_features, 1) or \
83.     (n_targets, n_features)
84.     ``sparse_coef_`` is a readonly property derived from ``coef_``
85.     
86.     intercept_ : float or ndarray of shape (n_targets,)
87.     independent term in decision function.
88.     
89.     n_iter_ : int or list of int
90.     number of iterations run by the coordinate descent solver to reach
91.     the specified tolerance.
92.     
93.     Examples
94.     --------
95.     >>> from sklearn import linear_model
96.     >>> clf = linear_model.Lasso(alpha=0.1)
97.     >>> clf.fit([[0,0], [1, 1], [2, 2]], [0, 1, 2])
98.     Lasso(alpha=0.1)
99.     >>> print(clf.coef_)
100.     [0.85 0.  ]
101.     >>> print(clf.intercept_)
102.     0.15...
103.     
104.     See also
105.     --------
106.     lars_path
107.     lasso_path
108.     LassoLars
109.     LassoCV
110.     LassoLarsCV
111.     sklearn.decomposition.sparse_encode
112.     
113.     Notes
114.     -----
115.     The algorithm used to fit the model is coordinate descent.
116.     
117.     To avoid unnecessary memory duplication the X argument of the fit 
118.      method
119.     should be directly passed as a Fortran-contiguous numpy array.
120.     """
121.     path = staticmethod(enet_path)
122.     @_deprecate_positional_args
123. def __init__(self, alpha=1.0, *, fit_intercept=True, normalize=False, 
124.         precompute=False, copy_X=True, max_iter=1000, 
125.         tol=1e-4, warm_start=False, positive=False, 
126.         random_state=None, selection='cyclic'):
127. super().__init__(alpha=alpha, l1_ratio=1.0, fit_intercept=fit_intercept, 
128.          normalize=normalize, precompute=precompute, copy_X=copy_X, 
129.          max_iter=max_iter, tol=tol, warm_start=warm_start, positive=positive, 
130.          random_state=random_state, selection=selection)
131. 
132. 
133. ######################################################
134. #########################
135. # Functions for CV with paths functions