In [22]:

    

from sklearn import preprocessing, datasets as d
import numpy as np
import scipy


    

In [8]:

    

boston = d.load_boston()
X, y = boston.data, boston.target


    

In [9]:

    

X[:, :3].mean(axis=0)


    

    
Out[9]:





array([  3.59376071,  11.36363636,  11.13677866])

In [10]:

    

X[:, :3].std(axis=0)


    

    
Out[10]:





array([  8.58828355,  23.29939569,   6.85357058])

In [11]:

    

X_2 = preprocessing.scale(X[:, :3])


    

In [12]:

    

X_2.mean(axis=0)


    

    
Out[12]:





array([  6.34099712e-17,  -6.34319123e-16,  -2.68291099e-15])

In [13]:

    

X_2.std(axis=0)


    

    
Out[13]:





array([ 1.,  1.,  1.])

In [14]:

    

normalized_X = preprocessing.normalize(X[:, :3])


    

In [16]:

    

normalized_X.mean(axis=0)


    

    
Out[16]:





array([ 0.14564668,  0.25834782,  0.73560531])

In [17]:

    

normalized_X.std(axis=0)


    

    
Out[17]:





array([ 0.22560452,  0.43091533,  0.36652811])

In [19]:

    

test_x = [[1.0,1.0,0.0], [3.0,3.0,0.0], [1.0,-1.0, 0.0]]
normalized_test = preprocessing.normalize(test_x)


    

In [20]:

    

normalized_test


    

    
Out[20]:





array([[ 0.70710678,  0.70710678,  0.        ],
       [ 0.70710678,  0.70710678,  0.        ],
       [ 0.70710678, -0.70710678,  0.        ]])

In [21]:

    

## handling sparse imputations


    

In [23]:

    

matrix = scipy.sparse.eye(1000)


    

In [25]:

    

preprocessing.scale(matrix, with_mean=False)


    

    
Out[25]:





<1000x1000 sparse matrix of type '<type 'numpy.float64'>'
	with 1000 stored elements in Compressed Sparse Row format>

In [ ]: