In [1]:

    

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline
from pandas import DataFrame, Series


    

In [2]:

    

arr = np.arange(3)

arr.repeat(3)


    

    
Out[2]:





array([0, 0, 0, 1, 1, 1, 2, 2, 2])

In [3]:

    

arr.repeat([2, 3, 4])


    

    
Out[3]:





array([0, 0, 1, 1, 1, 2, 2, 2, 2])

In [4]:

    

from numpy.random import randn


    

In [5]:

    

arr = randn(2, 2)
arr


    

    
Out[5]:





array([[ 1.63601648, -0.83344146],
       [ 0.47724003, -0.08775426]])

In [6]:

    

arr.repeat(2, axis=0)


    

    
Out[6]:





array([[ 1.63601648, -0.83344146],
       [ 1.63601648, -0.83344146],
       [ 0.47724003, -0.08775426],
       [ 0.47724003, -0.08775426]])

In [7]:

    

arr.repeat([2, 3], axis=0)


    

    
Out[7]:





array([[ 1.63601648, -0.83344146],
       [ 1.63601648, -0.83344146],
       [ 0.47724003, -0.08775426],
       [ 0.47724003, -0.08775426],
       [ 0.47724003, -0.08775426]])

In [8]:

    

arr.repeat([2, 3], axis=1)


    

    
Out[8]:





array([[ 1.63601648,  1.63601648, -0.83344146, -0.83344146, -0.83344146],
       [ 0.47724003,  0.47724003, -0.08775426, -0.08775426, -0.08775426]])

In [9]:

    

arr


    

    
Out[9]:





array([[ 1.63601648, -0.83344146],
       [ 0.47724003, -0.08775426]])

In [10]:

    

np.tile(arr, 2)


    

    
Out[10]:





array([[ 1.63601648, -0.83344146,  1.63601648, -0.83344146],
       [ 0.47724003, -0.08775426,  0.47724003, -0.08775426]])

In [11]:

    

np.tile(arr, (2, 1))


    

    
Out[11]:





array([[ 1.63601648, -0.83344146],
       [ 0.47724003, -0.08775426],
       [ 1.63601648, -0.83344146],
       [ 0.47724003, -0.08775426]])

In [12]:

    

np.tile(arr, (3, 2))


    

    
Out[12]:





array([[ 1.63601648, -0.83344146,  1.63601648, -0.83344146],
       [ 0.47724003, -0.08775426,  0.47724003, -0.08775426],
       [ 1.63601648, -0.83344146,  1.63601648, -0.83344146],
       [ 0.47724003, -0.08775426,  0.47724003, -0.08775426],
       [ 1.63601648, -0.83344146,  1.63601648, -0.83344146],
       [ 0.47724003, -0.08775426,  0.47724003, -0.08775426]])

In [13]:

    

## 花式索引的等价函数


    

In [14]:

    

arr = np.arange(10) * 100
inds = [7, 1, 2, 6]
arr[inds]


    

    
Out[14]:





array([700, 100, 200, 600])

In [15]:

    

arr.take(inds)


    

    
Out[15]:





array([700, 100, 200, 600])

In [17]:

    

arr.put(inds, 42)
arr


    

    
Out[17]:





array([  0,  42,  42, 300, 400, 500,  42,  42, 800, 900])

In [18]:

    

arr.put(inds, [40, 41, 42, 43])
arr


    

    
Out[18]:





array([  0,  41,  42, 300, 400, 500,  43,  40, 800, 900])

In [19]:

    

inds = [2, 0, 2, 1]
arr = randn(2, 4)
arr


    

    
Out[19]:





array([[ 0.93129879, -0.35411414,  0.57702374,  0.24116572],
       [-0.83636279,  0.91580841,  0.31046141,  1.75450793]])

In [20]:

    

arr.take(inds, axis=1)


    

    
Out[20]:





array([[ 0.57702374,  0.93129879,  0.57702374, -0.35411414],
       [ 0.31046141, -0.83636279,  0.31046141,  0.91580841]])

In [21]:

    

arr = randn(1000, 50)
# 500行随机样本
inds = np.random.permutation(1000)[:500]
%timeit arr[inds]


    

    




The slowest run took 4.15 times longer than the fastest. This could mean that an intermediate result is being cached.
10000 loops, best of 3: 34 µs per loop

In [22]:

    

%timeit arr.take(inds, axis=0)


    

    




10000 loops, best of 3: 25.9 µs per loop

In [23]:

    

arr = np.arange(5)
arr


    

    
Out[23]:





array([0, 1, 2, 3, 4])

In [24]:

    

arr * 4


    

    
Out[24]:





array([ 0,  4,  8, 12, 16])

In [30]:

    

arr = randn(4, 3)
arr.mean(0)


    

    
Out[30]:





array([ 0.19983727,  0.37342783, -0.05995232])

In [31]:

    

demeaned = arr - arr.mean(0)
demeaned


    

    
Out[31]:





array([[-0.27610684, -0.21123702, -0.16088269],
       [-0.84854155,  0.50728507,  0.07791194],
       [-0.27615565,  0.76700785, -0.69072193],
       [ 1.40080404, -1.0630559 ,  0.77369269]])

In [32]:

    

demeaned.mean(0)


    

    
Out[32]:





array([ -5.55111512e-17,   0.00000000e+00,   0.00000000e+00])

In [35]:

    

row_means = arr.mean(1)
row_means.reshape((4, 1))


    

    
Out[35]:





array([[-0.04497126],
       [ 0.08332275],
       [ 0.10448102],
       [ 0.54158454]])

In [36]:

    

demeaned = arr - row_means.reshape((4, 1))
demeaned.mean(1)


    

    
Out[36]:





array([ -9.25185854e-18,   3.23815049e-17,   0.00000000e+00,
        -3.70074342e-17])

In [37]:

    

arr - arr.mean(1)


    

    




---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-37-7b87b85a20b2> in <module>()
----> 1 arr - arr.mean(1)

ValueError: operands could not be broadcast together with shapes (4,3) (4,) 

In [38]:

    

arr - arr.mean(1).reshape((4, 1))


    

    
Out[38]:





array([[-0.03129831,  0.20716207, -0.17586375],
       [-0.73202702,  0.79739015, -0.06536313],
       [-0.1807994 ,  1.03595467, -0.85515527],
       [ 1.05905677, -1.2312126 ,  0.17215583]])

In [40]:

    

arr = np.zeros((4, 4))
arr_3d = arr[:, np.newaxis, :]
arr_3d.shape


    

    
Out[40]:





(4L, 1L, 4L)

In [41]:

    

arr_1d = np.random.normal(size=3)
arr_1d[:, np.newaxis]


    

    
Out[41]:





array([[ 0.31555779],
       [ 0.17220104],
       [ 1.97139767]])

In [42]:

    

arr_1d[np.newaxis, :]


    

    
Out[42]:





array([[ 0.31555779,  0.17220104,  1.97139767]])

In [43]:

    

arr = randn(3, 4, 5)
depth_means = arr.mean(2)
depth_means


    

    
Out[43]:





array([[ 0.17218435, -0.34911607,  0.58013683,  0.41087967],
       [-0.61084914, -0.37566525,  0.29728931,  0.55414718],
       [ 0.32010987,  0.15772747, -0.60769567, -0.57177377]])

In [44]:

    

demeaned = arr - depth_means[:, :, np.newaxis]
demeaned.mean(2)


    

    
Out[44]:





array([[  1.11022302e-17,  -8.88178420e-17,   4.44089210e-17,
          0.00000000e+00],
       [ -2.22044605e-17,  -3.33066907e-17,   0.00000000e+00,
          0.00000000e+00],
       [  2.22044605e-17,  -1.11022302e-17,   6.66133815e-17,
         -4.44089210e-17]])

In [ ]:

    

def demean_axis(arr, axis=0):
    means = arr.mean(axis)
    
    # 下面这些一般化的东西类似于N维的[:, :, np.newaxis]
    indexer = [slice(None)] * arr.ndim
    indexer[axis] = np.newaxis
    return arr - means[indexer]


    

In [45]:

    

arr = np.zeros((4, 3))
arr[:] = 5
arr


    

    
Out[45]:





array([[ 5.,  5.,  5.],
       [ 5.,  5.,  5.],
       [ 5.,  5.,  5.],
       [ 5.,  5.,  5.]])

In [46]:

    

col = np.array([1.28, -0.42, 0.44, 1.6])
arr[:] = col[:, np.newaxis]
arr


    

    
Out[46]:





array([[ 1.28,  1.28,  1.28],
       [-0.42, -0.42, -0.42],
       [ 0.44,  0.44,  0.44],
       [ 1.6 ,  1.6 ,  1.6 ]])

In [47]:

    

arr[:2] = [[-1.37], [0.509]]
arr


    

    
Out[47]:





array([[-1.37 , -1.37 , -1.37 ],
       [ 0.509,  0.509,  0.509],
       [ 0.44 ,  0.44 ,  0.44 ],
       [ 1.6  ,  1.6  ,  1.6  ]])

In [ ]: