diff --git a/scikits/learn/datasets/samples_generator.py b/scikits/learn/datasets/samples_generator.py
index dcac19f7af4757c7a463cb4b3f640f3d908e7606..be11dff5d86f869a00712ba26076fcb5e950de53 100644
--- a/scikits/learn/datasets/samples_generator.py
+++ b/scikits/learn/datasets/samples_generator.py
@@ -2,38 +2,36 @@
Generate samples of synthetic data sets.
"""
-# Author: B. Thirion, G. Varoquaux, A. Gramfort, V. Michel
+# Author: B. Thirion, G. Varoquaux, A. Gramfort, V. Michel, O. Grisel
# License: BSD 3 clause
import numpy as np
import numpy.random as nr
-
-def samples_classif():
- pass
-
-######################################################################
-# Generate Dataset for test
-######################################################################
-
def test_dataset_classif(n_samples=100, n_features=100, param=[1,1],
- k=0, seed=None):
- """
- Generate an snp matrix
+ n_informative=0, k=0, seed=None):
+ """Generate an snp matrix
Parameters
----------
n_samples : 100, int,
- the number of subjects
+ the number of observations
+
n_features : 100, int,
- the number of features
- param : [1,1], list,
+ the number of features for each observation
+
+ param : [1, 1], list,
parameter of a dirichlet density
that is used to generate multinomial densities
- from which the n_featuress will be samples
- k : 0, int,
+ from which the n_features will be samples
+
+ n_informative: 0, int
number of informative features
+
+ k : 0, int
+ deprecated: use n_informative instead
+
seed : None, int or np.random.RandomState
if seed is an instance of np.random.RandomState,
it is used to initialize the random generator
@@ -42,12 +40,18 @@ def test_dataset_classif(n_samples=100, n_features=100, param=[1,1],
-------
x : array of shape(n_samples, n_features),
the design matrix
+
y : array of shape (n_samples),
the subject labels
"""
- assert k<=n_features, ValueError('cannot have %d informative features and'
- ' %d features' % (k, n_features))
+ if k > 0 and n_informative == 0:
+ n_informative = k
+
+ if n_informative > n_features:
+ raise ValueError('cannot have %d informative features and'
+ ' %d features' % (n_informative, n_features))
+
if isinstance(seed, np.random.RandomState):
random = seed
elif seed is not None:
@@ -59,22 +63,29 @@ def test_dataset_classif(n_samples=100, n_features=100, param=[1,1],
y = np.zeros(n_samples)
param = np.ravel(np.array(param)).astype(np.float)
for n in range(n_samples):
- y[n] = np.nonzero(random.multinomial(1, param/param.sum()))[0]
- x[:,:k] += 3*y[:,np.newaxis]
+ y[n] = np.nonzero(random.multinomial(1, param / param.sum()))[0]
+ x[:, :k] += 3 * y[:, np.newaxis]
return x, y
-def test_dataset_reg(n_samples=100, n_features=100, k=0, seed=None):
- """
- Generate an snp matrix
+
+def test_dataset_reg(n_samples=100, n_features=100, n_informative=0, k=0,
+ seed=None):
+ """Generate an snp matrix
Parameters
----------
- n_samples : 100, int,
+ n_samples : 100, int
the number of subjects
- n_features : 100, int,
+
+ n_features : 100, int
the number of features
- k : 0, int,
+
+ n_informative: 0, int
number of informative features
+
+ k : 0, int
+ deprecated: use n_informative instead
+
seed : None, int or np.random.RandomState
if seed is an instance of np.random.RandomState,
it is used to initialize the random generator
@@ -83,12 +94,17 @@ def test_dataset_reg(n_samples=100, n_features=100, k=0, seed=None):
-------
x : array of shape(n_samples, n_features),
the design matrix
+
y : array of shape (n_samples),
the subject data
-
"""
- assert k<n_features, ValueError('cannot have %d informative fetaures and'
- ' %d features' % (k, n_features))
+ if k > 0 and n_informative == 0:
+ n_informative = k
+
+ if n_informative > n_features:
+ raise ValueError('cannot have %d informative features and'
+ ' %d features' % (n_informative, n_features))
+
if isinstance(seed, np.random.RandomState):
random = seed
elif seed is not None:
@@ -98,67 +114,68 @@ def test_dataset_reg(n_samples=100, n_features=100, k=0, seed=None):
x = random.randn(n_samples, n_features)
y = random.randn(n_samples)
- x[:,:k] += y[:, np.newaxis]
+ x[:, :k] += y[:, np.newaxis]
return x, y
+def sparse_uncorrelated(n_samples=100, n_features=10):
+ """Function creating simulated data with sparse uncorrelated design
+ cf.Celeux et al. 2009, Bayesian regularization in regression)
-
-######################################################################
-# Generate Dataset for regression
-######################################################################
-
-
-def sparse_uncorrelated(nb_samples=100, nb_features=10):
- """
- Function creating simulated data with sparse uncorrelated design.
- (cf.Celeux et al. 2009, Bayesian regularization in regression)
- X = NR.normal(0,1)
- Y = NR.normal(X[:,0]+2*X[:,1]-2*X[:,2]-1.5*X[:,3])
+ X = NR.normal(0, 1)
+ Y = NR.normal(X[:, 0] + 2 * X[:, 1] - 2 * X[:, 2] - 1.5 * X[:, 3])
The number of features is at least 10.
Parameters
----------
- nb_samples : int
- number of samples (default is 100).
- nb_features : int
- number of features (default is 5).
+ n_samples : int
+ number of samples (default is 100).
+ n_features : int
+ number of features (default is 10).
Returns
-------
- X : numpy array of shape (nb_samples, nb_features) for input samples
- Y : numpy array of shape (nb_samples) for labels
+ X : numpy array of shape (n_samples, n_features) for input samples
+ y : numpy array of shape (n_samples) for labels
"""
- X = nr.normal(loc=0, scale=1, size=(nb_samples, nb_features))
- Y = nr.normal(loc=X[:, 0] + 2 * X[:, 1] - 2 * X[:,2] - 1.5 * X[:, 3],
- scale = np.ones(nb_samples))
- return X, Y
+ X = nr.normal(loc=0, scale=1, size=(n_samples, n_features))
+ y = nr.normal(loc=X[:, 0] + 2 * X[:, 1] - 2 * X[:,2] - 1.5 * X[:, 3],
+ scale = np.ones(n_samples))
+ return X, y
-def friedman(nb_samples=100, nb_features=10,noise_std=1):
- """
- Function creating simulated data with non linearities
- (cf.Friedman 1993)
- X = NR.normal(0,1)
- Y = 10*sin(X[:,0]*X[:,1]) + 20*(X[:,2]-0.5)**2 + 10*X[:,3] + 5*X[:,4]
+def friedman(n_samples=100, n_features=10, noise_std=1):
+ """Function creating simulated data with non linearities
+
+ cf. Friedman 1993
+
+ X = np.random.normal(0, 1)
+
+ y = 10 * sin(X[:, 0] * X[:, 1]) + 20 * (X[:, 2] - 0.5) ** 2 \
+ + 10 * X[:, 3] + 5 * X[:, 4]
+
The number of features is at least 5.
Parameters
----------
- nb_samples : int
- number of samples (default is 100).
- nb_features : int
- number of features (default is 10).
+ n_samples : int
+ number of samples (default is 100).
+
+ n_features : int
+ number of features (default is 10).
+
noise_std : float
- std of the noise, which is added as noise_std*NR.normal(0,1)
+ std of the noise, which is added as noise_std*NR.normal(0,1)
+
Returns
-------
- X : numpy array of shape (nb_samples, nb_features) for input samples
- Y : numpy array of shape (nb_samples) for labels
-
+ X : numpy array of shape (n_samples, n_features) for input samples
+ y : numpy array of shape (n_samples,) for labels
"""
- X = nr.normal(loc=0, scale=1, size=(nb_samples, nb_features))
- Y = 10*np.sin(X[:,0]*X[:,1]) + 20*(X[:,2]-0.5)**2 + 10*X[:,3] + 5*X[:,4]
- Y += noise_std*nr.normal(loc=0,scale=1,size=(nb_samples))
- return X,Y
+ X = nr.normal(loc=0, scale=1, size=(n_samples, n_features))
+ y = 10 * np.sin(X[:, 0] * X[:, 1]) + 20 * (X[:, 2] - 0.5) ** 2 \
+ + 10 * X[:, 3] + 5 * X[:, 4]
+ y += noise_std * nr.normal(loc=0, scale=1, size=n_samples)
+ return X, y
+