scikit-learn/sklearn/linear_model/stochastic_gradient.py

757 lines
28 KiB
Python

# Authors: Peter Prettenhofer <peter.prettenhofer@gmail.com> (main author)
# Mathieu Blondel (partial_fit support)
#
# License: BSD Style.
"""Implementation of Stochastic Gradient Descent (SGD)."""
import numpy as np
import scipy.sparse as sp
import warnings
from ..externals.joblib import Parallel, delayed
from ..base import RegressorMixin
from ..base import ClassifierMixin
from ..feature_selection.selector_mixin import SelectorMixin
from .base import BaseSGD
from ..utils import atleast2d_or_csr, check_arrays
from ..utils.extmath import safe_sparse_dot
from ..utils import safe_asarray
from ..utils import deprecated
from .sgd_fast import plain_sgd as plain_sgd
from ..utils.seq_dataset import ArrayDataset, CSRDataset
from .sgd_fast import Hinge, Log, ModifiedHuber, SquaredLoss, Huber
def _make_dataset(X, y_i, sample_weight):
"""Returns Dataset object + intercept_decay"""
if sp.issparse(X):
dataset = CSRDataset(X.data, X.indptr, X.indices, y_i, sample_weight)
intercept_decay = 0.01
else:
dataset = ArrayDataset(X, y_i, sample_weight)
intercept_decay = 1.0
return dataset, intercept_decay
def _tocsr(X):
"""Convert X to CSR matrix, preventing a copy if possible"""
if sp.isspmatrix_csr(X) and X.dtype == np.float64:
return X
else:
return sp.csr_matrix(X, dtype=np.float64)
class SGDClassifier(BaseSGD, ClassifierMixin, SelectorMixin):
"""Linear model fitted by minimizing a regularized empirical loss with SGD.
SGD stands for Stochastic Gradient Descent: 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).
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.
This implementation works with data represented as dense numpy arrays of
floating point values for the features.
Parameters
----------
loss : str, 'hinge' or 'log' or 'modified_huber'
The loss function to be used. Defaults to 'hinge'. The hinge loss is
a margin loss used by standard linear SVM models. The 'log' loss is
the loss of logistic regression models and can be used for
probability estimation in binary classifiers. 'modified_huber'
is another smooth loss that brings tolerance to outliers.
penalty : str, 'l2' or 'l1' or 'elasticnet'
The penalty (aka regularization term) to be used. Defaults to 'l2'
which is the standard regularizer for linear SVM models. 'l1' and
'elasticnet' migh bring sparsity to the model (feature selection)
not achievable with 'l2'.
alpha : float
Constant that multiplies the regularization term. Defaults to 0.0001
rho : float
The Elastic Net mixing parameter, with 0 < rho <= 1.
Defaults to 0.85.
fit_intercept: bool
Whether the intercept should be estimated or not. If False, the
data is assumed to be already centered. Defaults to True.
n_iter: int, optional
The number of passes over the training data (aka epochs).
Defaults to 5.
shuffle: bool, optional
Whether or not the training data should be shuffled after each epoch.
Defaults to False.
seed: int, optional
The seed of the pseudo random number generator to use when
shuffling the data.
verbose: integer, optional
The verbosity level
n_jobs: integer, optional
The number of CPUs to use to do the OVA (One Versus All, for
multi-class problems) computation. -1 means 'all CPUs'. Defaults
to 1.
learning_rate : string, optional
The learning rate:
constant: eta = eta0
optimal: eta = 1.0/(t+t0) [default]
invscaling: eta = eta0 / pow(t, power_t)
eta0 : double
The initial learning rate [default 0.01].
power_t : double
The exponent for inverse scaling learning rate [default 0.25].
class_weight : dict, {class_label : weight} or "auto" or None, optional
Preset for the class_weight fit parameter.
Weights associated with classes. If not given, all classes
are supposed to have weight one.
The "auto" mode uses the values of y to automatically adjust
weights inversely proportional to class frequencies.
warm_start : bool, optional
When set to True, reuse the solution of the previous call to fit as
initialization, otherwise, just erase the previous solution.
Attributes
----------
`coef_` : array, shape = [1, n_features] if n_classes == 2 else [n_classes,
n_features]
Weights assigned to the features.
`intercept_` : array, shape = [1] if n_classes == 2 else [n_classes]
Constants in decision function.
Examples
--------
>>> import numpy as np
>>> from sklearn import linear_model
>>> X = np.array([[-1, -1], [-2, -1], [1, 1], [2, 1]])
>>> Y = np.array([1, 1, 2, 2])
>>> clf = linear_model.SGDClassifier()
>>> clf.fit(X, Y)
... #doctest: +NORMALIZE_WHITESPACE
SGDClassifier(alpha=0.0001, class_weight=None, eta0=0.0,
fit_intercept=True, learning_rate='optimal', loss='hinge',
n_iter=5, n_jobs=1, penalty='l2', power_t=0.5, rho=0.85, seed=0,
shuffle=False, verbose=0, warm_start=False)
>>> print clf.predict([[-0.8, -1]])
[1]
See also
--------
LinearSVC, LogisticRegression, Perceptron
"""
def __init__(self, loss="hinge", penalty='l2', alpha=0.0001,
rho=0.85, fit_intercept=True, n_iter=5, shuffle=False,
verbose=0, n_jobs=1, seed=0, learning_rate="optimal",
eta0=0.0, power_t=0.5, class_weight=None, warm_start=False):
super(SGDClassifier, self).__init__(loss=loss, penalty=penalty,
alpha=alpha, rho=rho,
fit_intercept=fit_intercept,
n_iter=n_iter, shuffle=shuffle,
verbose=verbose, seed=seed,
learning_rate=learning_rate,
eta0=eta0, power_t=power_t,
warm_start=warm_start)
self.class_weight = class_weight
self.classes_ = None
self.n_jobs = int(n_jobs)
@property
@deprecated("to be removed in v0.12; use ``classes_`` instead.")
def classes(self):
return self.classes_
def _set_loss_function(self, loss):
"""Set concrete LossFunction."""
loss_functions = {
"hinge": Hinge(1.0),
"perceptron": Hinge(0.0),
"log": Log(),
"modified_huber": ModifiedHuber(),
}
try:
self.loss_function = loss_functions[loss]
except KeyError:
raise ValueError("The loss %s is not supported. " % loss)
def _set_class_weight(self, class_weight, classes, y):
"""Estimate class weights for unbalanced datasets."""
if class_weight is None:
# keep the old class_weight if none provided
class_weight = self.class_weight
if class_weight is None or len(class_weight) == 0:
# uniform class weights
weight = np.ones(classes.shape[0], dtype=np.float64, order='C')
elif class_weight == 'auto':
# proportional to the number of samples in the class
weight = np.array([1.0 / np.sum(y == i) for i in classes],
dtype=np.float64, order='C')
weight *= classes.shape[0] / np.sum(weight)
else:
# user-defined dictionary
weight = np.ones(classes.shape[0], dtype=np.float64, order='C')
if not isinstance(class_weight, dict):
raise ValueError("class_weight must be dict, 'auto', or None,"
" got: %r" % class_weight)
for c in class_weight:
i = np.searchsorted(classes, c)
if classes[i] != c:
raise ValueError("Class label %d not present." % c)
else:
weight[i] = class_weight[c]
self._expanded_class_weight = weight
def _partial_fit(self, X, y, n_iter, classes=None, sample_weight=None):
X = safe_asarray(X, dtype=np.float64, order="C")
y = np.asarray(y)
n_samples, n_features = X.shape
self._check_fit_data(X, y)
if self.classes_ is None and classes is None:
raise ValueError("classes must be passed on the first call "
"to partial_fit.")
elif classes is not None and self.classes_ is not None:
if not np.all(self.classes_ == np.unique(classes)):
raise ValueError("`classes` is not the same as on last call "
"to partial_fit.")
elif classes is not None:
self.classes_ = classes
n_classes = self.classes_.shape[0]
# Allocate datastructures from input arguments
self._set_class_weight(self.class_weight, self.classes_, y)
sample_weight = self._validate_sample_weight(sample_weight, n_samples)
if self.coef_ is None:
self._allocate_parameter_mem(n_classes, n_features,
coef_init=None, intercept_init=None)
# delegate to concrete training procedure
if n_classes > 2:
self._fit_multiclass(X, y, sample_weight, n_iter)
elif n_classes == 2:
self._fit_binary(X, y, sample_weight, n_iter)
else:
raise ValueError("The number of class labels must be "
"greater than one.")
self.t_ += n_iter * n_samples
return self
def partial_fit(self, X, y, classes=None,
class_weight=None, sample_weight=None):
"""Fit linear model with Stochastic Gradient Descent.
Parameters
----------
X : {array-like, sparse matrix}, shape = [n_samples, n_features]
Subset of the training data
y : numpy array of shape [n_samples]
Subset of the target values
classes : array, shape = [n_classes]
Classes across all calls to partial_fit.
Can be obtained by via `np.unique(y_all)`, where y_all is the
target vector of the entire dataset.
This argument is required for the first call to partial_fit
and can be omitted in the subsequent calls.
Note that y doesn't need to contain all labels in `classes`.
sample_weight : array-like, shape = [n_samples], optional
Weights applied to individual samples.
If not provided, uniform weights are assumed.
Returns
-------
self : returns an instance of self.
"""
if class_weight != None:
warnings.warn("Using 'class_weight' as a parameter to the 'fit'"
"method is deprecated. Set it on initialization instead.",
DeprecationWarning)
self.class_weight = class_weight
return self._partial_fit(X, y, n_iter=1, classes=classes,
sample_weight=sample_weight)
def fit(self, X, y, coef_init=None, intercept_init=None,
class_weight=None, sample_weight=None):
"""Fit linear model with Stochastic Gradient Descent.
Parameters
----------
X : {array-like, sparse matrix}, shape = [n_samples, n_features]
Training data
y : numpy array of shape [n_samples]
Target values
coef_init : array, shape = [n_classes,n_features]
The initial coeffients to warm-start the optimization.
intercept_init : array, shape = [n_classes]
The initial intercept 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.
Returns
-------
self : returns an instance of self.
"""
if class_weight != None:
warnings.warn("Using 'class_weight' as a parameter to the 'fit'"
"method is deprecated. Set it on initialization instead.",
DeprecationWarning)
self.class_weight = class_weight
X = safe_asarray(X, dtype=np.float64, order="C")
y = np.asarray(y)
n_samples, n_features = X.shape
self._check_fit_data(X, y)
# np.unique sorts in asc order; largest class id is positive class
classes = np.unique(y)
n_classes = classes.shape[0]
if self.warm_start and self.coef_ is not None:
if coef_init is None:
coef_init = self.coef_
if intercept_init is None:
intercept_init = self.intercept_
# Allocate datastructures from input arguments.
self._allocate_parameter_mem(n_classes, n_features,
coef_init, intercept_init)
# Need to re-initialize in case of multiple call to fit.
self._init_t()
self._partial_fit(X, y, self.n_iter,
classes=classes,
sample_weight=sample_weight)
# fitting is over, we can now transform coef_ to fortran order
# for faster predictions
self._set_coef(self.coef_)
return self
def decision_function(self, X):
"""Predict signed 'distance' to the hyperplane (aka confidence score)
Parameters
----------
X : {array-like, sparse matrix}, shape = [n_samples, n_features]
Returns
-------
array, shape = [n_samples] if n_classes == 2 else [n_samples,n_classes]
The signed 'distances' to the hyperplane(s).
"""
X = atleast2d_or_csr(X)
scores = safe_sparse_dot(X, self.coef_.T) + self.intercept_
if self.classes_.shape[0] == 2:
return np.ravel(scores)
else:
return scores
def predict(self, X):
"""Predict using the linear model
Parameters
----------
X : {array-like, sparse matrix}, shape = [n_samples, n_features]
Returns
-------
array, shape = [n_samples]
Array containing the predicted class labels.
"""
scores = self.decision_function(X)
if self.classes_.shape[0] == 2:
indices = np.array(scores > 0, dtype=np.int)
else:
indices = scores.argmax(axis=1)
return self.classes_[np.ravel(indices)]
def predict_proba(self, X):
"""Predict class membership probability
Parameters
----------
X : {array-like, sparse matrix}, shape = [n_samples, n_features]
Returns
-------
array, shape = [n_samples] if n_classes == 2 else [n_samples,
n_classes]
Contains the membership probabilities of the positive class.
"""
if len(self.classes_) != 2:
raise NotImplementedError("predict_(log_)proba only supported"
" for binary classification")
elif not isinstance(self.loss_function, Log):
raise NotImplementedError("predict_(log_)proba only supported when"
" loss='log' (%s given)" % self.loss)
return 1.0 / (1.0 + np.exp(-self.decision_function(X)))
def _fit_binary(self, X, y, sample_weight, n_iter):
if sp.issparse(X):
X = _tocsr(X)
coef, intercept = fit_binary(self, 1, X, y, n_iter,
self._expanded_class_weight[1],
self._expanded_class_weight[0],
sample_weight)
# need to be 2d
self.coef_ = coef.reshape(1, -1)
# intercept is a float, need to convert it to an array of length 1
self.intercept_ = np.atleast_1d(intercept)
def _fit_multiclass(self, X, y, sample_weight, n_iter):
"""Fit 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.
"""
if sp.issparse(X):
X = _tocsr(X)
# Use joblib to fit OvA in parallel
result = Parallel(n_jobs=self.n_jobs, verbose=self.verbose)(
delayed(fit_binary)(self, i, X, y, n_iter,
self._expanded_class_weight[i], 1.,
sample_weight)
for i in xrange(len(self.classes_)))
for i, (coef, intercept) in enumerate(result):
self.coef_[i] = coef
self.intercept_[i] = intercept
def _prepare_fit_binary(est, y, i):
"""Common initialization for _fit_binary_{dense,sparse}.
Returns y, coef, intercept.
"""
y_i = np.ones(y.shape, dtype=np.float64, order="C")
y_i[y != est.classes_[i]] = -1.0
if len(est.classes_) == 2:
coef = est.coef_.ravel()
intercept = est.intercept_[0]
else:
coef = est.coef_[i]
intercept = est.intercept_[i]
return y_i, coef, intercept
def fit_binary(est, i, X, y, n_iter, pos_weight, neg_weight,
sample_weight):
"""Fit a single binary classifier.
The i'th class is considered the "positive" class.
"""
y_i, coef, intercept = _prepare_fit_binary(est, y, i)
assert y_i.shape[0] == y.shape[0] == sample_weight.shape[0]
dataset, intercept_decay = _make_dataset(X, y_i, sample_weight)
return plain_sgd(coef, intercept, est.loss_function,
est.penalty_type, est.alpha, est.rho,
dataset, n_iter, est.fit_intercept,
est.verbose, est.shuffle, est.seed,
pos_weight, neg_weight,
est.learning_rate_code, est.eta0,
est.power_t, est.t_, intercept_decay)
class SGDRegressor(BaseSGD, RegressorMixin, SelectorMixin):
"""Linear model fitted by minimizing a regularized empirical loss with SGD
SGD stands for Stochastic Gradient Descent: 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).
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.
This implementation works with data represented as dense numpy arrays of
floating point values for the features.
Parameters
----------
loss : str, 'squared_loss' or 'huber'
The loss function to be used. Defaults to 'squared_loss' which refers
to the ordinary least squares fit. 'huber' is an epsilon insensitive
loss function for robust regression.
penalty : str, 'l2' or 'l1' or 'elasticnet'
The penalty (aka regularization term) to be used. Defaults to 'l2'
which is the standard regularizer for linear SVM models. 'l1' and
'elasticnet' migh bring sparsity to the model (feature selection)
not achievable with 'l2'.
alpha : float
Constant that multiplies the regularization term. Defaults to 0.0001
rho : float
The Elastic Net mixing parameter, with 0 < rho <= 1.
Defaults to 0.85.
fit_intercept: bool
Whether the intercept should be estimated or not. If False, the
data is assumed to be already centered. Defaults to True.
n_iter: int, optional
The number of passes over the training data (aka epochs).
Defaults to 5.
shuffle: bool, optional
Whether or not the training data should be shuffled after each epoch.
Defaults to False.
seed: int, optional
The seed of the pseudo random number generator to use when
shuffling the data.
verbose: integer, optional
The verbosity level.
p : float
Epsilon in the epsilon-insensitive huber loss function;
only if `loss=='huber'`.
learning_rate : string, optional
The learning rate:
constant: eta = eta0
optimal: eta = 1.0/(t+t0)
invscaling: eta = eta0 / pow(t, power_t) [default]
eta0 : double, optional
The initial learning rate [default 0.01].
power_t : double, optional
The exponent for inverse scaling learning rate [default 0.25].
warm_start : bool, optional
When set to True, reuse the solution of the previous call to fit as
initialization, otherwise, just erase the previous solution.
Attributes
----------
`coef_` : array, shape = [n_features]
Weights asigned to the features.
`intercept_` : array, shape = [1]
The intercept term.
Examples
--------
>>> import numpy as np
>>> from sklearn import linear_model
>>> n_samples, n_features = 10, 5
>>> np.random.seed(0)
>>> y = np.random.randn(n_samples)
>>> X = np.random.randn(n_samples, n_features)
>>> clf = linear_model.SGDRegressor()
>>> clf.fit(X, y)
SGDRegressor(alpha=0.0001, eta0=0.01, fit_intercept=True,
learning_rate='invscaling', loss='squared_loss', n_iter=5, p=0.1,
penalty='l2', power_t=0.25, rho=0.85, seed=0, shuffle=False,
verbose=0, warm_start=False)
See also
--------
Ridge, ElasticNet, Lasso, SVR
"""
def __init__(self, loss="squared_loss", penalty="l2", alpha=0.0001,
rho=0.85, fit_intercept=True, n_iter=5, shuffle=False,
verbose=0, p=0.1, seed=0, learning_rate="invscaling",
eta0=0.01, power_t=0.25, warm_start=False):
self.p = float(p)
super(SGDRegressor, self).__init__(loss=loss, penalty=penalty,
alpha=alpha, rho=rho,
fit_intercept=fit_intercept,
n_iter=n_iter, shuffle=shuffle,
verbose=verbose, seed=seed,
learning_rate=learning_rate,
eta0=eta0, power_t=power_t,
warm_start=False)
def _set_loss_function(self, loss):
"""Get concrete LossFunction"""
loss_functions = {
"squared_loss": SquaredLoss(),
"huber": Huber(self.p),
}
try:
self.loss_function = loss_functions[loss]
except KeyError:
raise ValueError("The loss %s is not supported. " % loss)
def _partial_fit(self, X, y, n_iter, sample_weight=None,
coef_init=None, intercept_init=None):
X, y = check_arrays(X, y, sparse_format="csr", copy=False,
check_ccontiguous=True, dtype=np.float64)
y = np.asarray(y, dtype=np.float64, order="C")
n_samples, n_features = X.shape
self._check_fit_data(X, y)
# Allocate datastructures from input arguments
sample_weight = self._validate_sample_weight(sample_weight, n_samples)
if self.coef_ is None:
self._allocate_parameter_mem(1, n_features,
coef_init, intercept_init)
self._fit_regressor(X, y, sample_weight, n_iter)
self.t_ += n_iter * n_samples
return self
def partial_fit(self, X, y, sample_weight=None):
"""Fit linear model with Stochastic Gradient Descent.
Parameters
----------
X : {array-like, sparse matrix}, shape = [n_samples, n_features]
Subset of training data
y : numpy array of shape [n_samples]
Subset of target values
sample_weight : array-like, shape = [n_samples], optional
Weights applied to individual samples.
If not provided, uniform weights are assumed.
Returns
-------
self : returns an instance of self.
"""
return self._partial_fit(X, y, n_iter=1, sample_weight=sample_weight)
def fit(self, X, y, coef_init=None, intercept_init=None,
sample_weight=None):
"""Fit linear model with Stochastic Gradient Descent.
Parameters
----------
X : {array-like, sparse matrix}, shape = [n_samples, n_features]
Training data
y : numpy array of shape [n_samples]
Target values
coef_init : array, shape = [n_features]
The initial coeffients to warm-start the optimization.
intercept_init : array, shape = [1]
The initial intercept to warm-start the optimization.
sample_weight : array-like, shape = [n_samples], optional
Weights applied to individual samples (1. for unweighted).
Returns
-------
self : returns an instance of self.
"""
if self.warm_start and self.coef_ is not None:
if coef_init is None:
coef_init = self.coef_
if intercept_init is None:
intercept_init = self.intercept_
# Need to re-initialize in case of multiple call to fit.
self._init_t()
return self._partial_fit(X, y, self.n_iter, sample_weight,
coef_init, intercept_init)
def decision_function(self, X):
"""Predict using the linear model
Parameters
----------
X : {array-like, sparse matrix}, shape = [n_samples, n_features]
Returns
-------
array, shape = [n_samples]
Predicted target values per element in X.
"""
X = atleast2d_or_csr(X)
scores = safe_sparse_dot(X, self.coef_) + self.intercept_
return scores.ravel()
def predict(self, X):
"""Predict using the linear model
Parameters
----------
X : {array-like, sparse matrix}, shape = [n_samples, n_features]
Returns
-------
array, shape = [n_samples]
Predicted target values per element in X.
"""
return self.decision_function(X)
def _fit_regressor(self, X, y, sample_weight, n_iter):
dataset, intercept_decay = _make_dataset(X, y, sample_weight)
self.coef_, intercept = plain_sgd(self.coef_,
self.intercept_[0],
self.loss_function,
self.penalty_type,
self.alpha, self.rho,
dataset,
n_iter,
int(self.fit_intercept),
int(self.verbose),
int(self.shuffle),
self.seed,
1.0, 1.0,
self.learning_rate_code,
self.eta0, self.power_t, self.t_,
intercept_decay)
self.intercept_ = np.atleast_1d(intercept)