1.Numpy是什么
非常easy。Numpy是Python的一个科学计算的库。提供了矩阵运算的功能,其一般与Scipy、matplotlib一起使用。事实上,list已经提供了类似于矩阵的表示形式,只是numpy为我们提供了很多其它的函数。
假设接触过matlab、scilab。那么numpy非常好入手。
在下面的代码演示样例中。总是先导入了numpy:(通用做法import numpu as np 简单输入)
>>> import numpy as np >>> print np.version.version 1.6.2
2. 多维数组
多维数组的类型是:numpy.ndarray。
使用numpy.array方法
以list或tuple变量为參数产生一维数组:
>>> print np.array([1,2,3,4]) [1 2 3 4] >>> print np.array((1.2,2,3,4)) [ 1.2 2. 3. 4. ] >>> print type(np.array((1.2,2,3,4))) <type 'numpy.ndarray'>
以list或tuple变量为元素产生二维数组或者多维数组:
>>> x = np.array(((1,2,3),(4,5,6)))
>>> x
array([[1, 2, 3],
[4, 5, 6]])
>>> y = np.array([[1,2,3],[4,5,6]])
>>> y
array([[1, 2, 3],
[4, 5, 6]])
numpy ndarray数据类型能够通过參数dtype 设定。并且能够使用astype转换类型。在处理文件时候这个会非常有用。注意astype 调用会返回一个新的数组,也就是原始数据的一份拷贝。
numeric_strings2 = np.array(['1.23','2.34','3.45'],dtype=np.string_)
numeric_strings2
Out[32]:
array(['1.23', '2.34', '3.45'],
dtype='|S4')
numeric_strings2.astype(float)
Out[33]: array([ 1.23, 2.34, 3.45])index 和slicing :第一数值类似数组横坐标。第二个为纵坐标
>>> x[1,2] 6 >>> y=x[:,1] >>> y array([2, 5])
涉及改变相关问题,我们改变上面y是否会改变x?这是特别须要关注的。
>>> y
array([2, 5])
>>> y[0] = 10
>>> y
array([10, 5])
>>> x
array([[ 1, 10, 3],
[ 4, 5, 6]])
通过上面能够发现改变y会改变x ,因而我们能够判断,y和x指向是同一块内存空间值,系统没有为y 新开辟空间把x值赋值过去。
arr = np.arange(10) arr Out[45]: array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9]) arr[4] Out[46]: 4 arr[3:6] Out[47]: array([3, 4, 5]) arr[3:6] = 12 arr Out[49]: array([ 0, 1, 2, 12, 12, 12, 6, 7, 8, 9])
思考为什么这么设计? Numpy 设计是为了处理大数据,假设切片採用数据复制话会产生极大的性能和内存消耗问题。
假如说须要对数组是一份副本而不是视图能够例如以下操作:
arr_copy = arr[3:6].copy() arr_copy[:]=24 arr_copy Out[54]: array([24, 24, 24]) arr Out[55]: array([ 0, 1, 2, 12, 12, 12, 6, 7, 8, 9])
再看下对list 切片改动
l=range(10)
l
Out[35]: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
l[5:8] = 12
Traceback (most recent call last):
File "<ipython-input-36-022af3ddcc9b>", line 1, in <module>
l[5:8] = 12
TypeError: can only assign an iterable
l1= l[5:8]
l1
Out[38]: [5, 6, 7]
l1[0]=12
l1
Out[40]: [12, 6, 7]
l
Out[41]: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]这里设计到python 中深浅拷贝,当中切片属于浅拷贝,详细參考:python深浅拷贝arr2d = np.arange(1,10).reshape(3,3)
arr2d
Out[57]:
array([[1, 2, 3],
[4, 5, 6],
[7, 8, 9]])
arr2d[2]
Out[58]: array([7, 8, 9])
arr2d[0][2]
Out[59]: 3
arr2d[0,2]
Out[60]: 3这样的类型在实际代码中出现比較多,关注下。
names = np.array(['Bob','joe','Bob','will']) names == 'Bob' Out[70]: array([ True, False, True, False], dtype=bool)
data
Out[73]:
array([[ 0.36762706, -1.55668952, 0.84316735, -0.116842 ],
[ 1.34023966, 1.12766186, 1.12507441, -0.68689309],
[ 1.27392366, -0.43399617, -0.80444728, 1.60731881],
[ 0.23361565, 1.38772715, 0.69129479, -1.19228023],
[ 0.51353082, 0.17696698, -0.06753478, 0.80448168],
[ 0.21773096, 0.60582802, -0.46446071, 0.83131122],
[ 0.50569072, 0.04431685, -0.69358155, -0.9629124 ]])
data[data < 0] = 0
data
Out[75]:
array([[ 0.36762706, 0. , 0.84316735, 0. ],
[ 1.34023966, 1.12766186, 1.12507441, 0. ],
[ 1.27392366, 0. , 0. , 1.60731881],
[ 0.23361565, 1.38772715, 0.69129479, 0. ],
[ 0.51353082, 0.17696698, 0. , 0.80448168],
[ 0.21773096, 0.60582802, 0. , 0.83131122],
[ 0.50569072, 0.04431685, 0. , 0. ]])上面展示通过布尔值来设置值的手段。
在跑实验时常常须要用到读取文件里的数据,事实上在numpy中已经有成熟函数封装好了能够使用
将数组以二进制形式格式保存到磁盘。np.save 、np.load 函数是读写磁盘的两个主要函数。默认情况下,数组以未压缩的原始二进制格式保存在扩展名为.npy的文件里
arr = np.arange(10)
np.save('some_array',arr)np.load('some_array.npy')
Out[80]: array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])存取文本文件:
文本中存放是聚类须要数据,直接能够方便读取到numpy array中,省去一行行读文件繁琐。
arr = np.loadtxt('dataMatrix.txt',delimiter=' ')
arr
Out[82]:
array([[ 1. , 1. , 1. , 1. , 1. ,
0.8125 ],
[ 0.52882353, 0.56271186, 0.48220588, 0.53384615, 0.61651376,
0.58285714],
[ 0. , 0. , 0. , 1. , 1. ,
1. ],
[ 1. , 0.92857143, 0.91857143, 1. , 1. ,
1. ],
[ 1. , 1. , 1. , 1. , 1. ,
1. ],
[ 0.05285714, 0.10304348, 0.068 , 0.06512821, 0.05492308,
0.05244898],
[ 0.04803279, 0.08203125, 0.05516667, 0.05517241, 0.04953488,
0.05591549],
[ 0.04803279, 0.08203125, 0.05516667, 0.05517241, 0.04953488,
0.05591549]])
使用numpy.arange方法
>>> print np.arange(15) [ 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14] >>> print type(np.arange(15)) <type 'numpy.ndarray'> >>> print np.arange(15).reshape(3,5) [[ 0 1 2 3 4] [ 5 6 7 8 9] [10 11 12 13 14]] >>> print type(np.arange(15).reshape(3,5)) <type 'numpy.ndarray'>
使用numpy.linspace方法
比如,在从1到10中产生20个数:
>>> print np.linspace(1,10,20) [ 1. 1.47368421 1.94736842 2.42105263 2.89473684 3.36842105 3.84210526 4.31578947 4.78947368 5.26315789 5.73684211 6.21052632 6.68421053 7.15789474 7.63157895 8.10526316 8.57894737 9.05263158 9.52631579 10. ]
使用numpy.zeros,numpy.ones。numpy.eye等方法能够构造特定的矩阵
>>> print np.zeros((3,4)) [[ 0. 0. 0. 0.] [ 0. 0. 0. 0.] [ 0. 0. 0. 0.]] >>> print np.ones((3,4)) [[ 1. 1. 1. 1.] [ 1. 1. 1. 1.] [ 1. 1. 1. 1.]] >>> print np.eye(3) [[ 1. 0. 0.] [ 0. 1. 0.] [ 0. 0. 1.]]
获取数组的属性:
>>> a = np.zeros((2,2,2)) >>> print a.ndim #数组的维数 3 >>> print a.shape #数组每一维的大小 (2, 2, 2) >>> print a.size #数组的元素数 8 >>> print a.dtype #元素类型 float64 >>> print a.itemsize #每一个元素所占的字节数 8
The following attributes contain information about the memory layout of the array:
| ndarray.flags | Information about the memory layout of the array. |
| ndarray.shape | Tuple of array dimensions. |
| ndarray.strides | Tuple of bytes to step in each dimension when traversing an array. |
| ndarray.ndim | Number of array dimensions. |
| ndarray.data | Python buffer object pointing to the start of the array’s data. |
| ndarray.size | Number of elements in the array. |
| ndarray.itemsize | Length of one array element in bytes. |
| ndarray.nbytes | Total bytes consumed by the elements of the array. |
| ndarray.base | Base object if memory is from some other object. |
An ndarray object has many methods which operate on or with the array in some fashion, typically returning an array result. These methods are briefly explained below. (Each method’s docstring has a more complete description.)
For the following methods there are also corresponding functions in numpy: all, any, argmax, argmin, argpartition, argsort, choose, clip,compress, copy, cumprod, cumsum, diagonal, imag, max, mean, min, nonzero, partition, prod, ptp, put, ravel, real, repeat, reshape, round,searchsorted, sort, squeeze, std, sum, swapaxes, take, trace, transpose, var.
很多其它Array的相关方法见:http://docs.scipy.org/doc/numpy/reference/arrays.ndarray.html
用到比較多函数演示样例:
>>> x
array([[[ 0, 1, 2],
[ 3, 4, 5],
[ 6, 7, 8]],
[[ 9, 10, 11],
[12, 13, 14],
[15, 16, 17]],
[[18, 19, 20],
[21, 22, 23],
[24, 25, 26]]])
>>> x.sum(axis=1)
array([[ 9, 12, 15],
[36, 39, 42],
[63, 66, 69]])
>>> x.sum(axis=2)
array([[ 3, 12, 21],
[30, 39, 48],
[57, 66, 75]])>>> np.sum([[0, 1], [0, 5]]) 6 >>> np.sum([[0, 1], [0, 5]], axis=0) array([0, 6]) >>> np.sum([[0, 1], [0, 5]], axis=1) array([1, 5])
合并数组
使用numpy下的vstack(垂直方向)和hstack(水平方向)函数:
>>> a = np.ones((2,2)) >>> b = np.eye(2) >>> print np.vstack((a,b)) [[ 1. 1.] [ 1. 1.] [ 1. 0.] [ 0. 1.]] >>> print np.hstack((a,b)) [[ 1. 1. 1. 0.] [ 1. 1. 0. 1.]]
看一下这两个函数有没有涉及到浅拷贝这样的问题:
>>> c = np.hstack((a,b)) >>> print c [[ 1. 1. 1. 0.] [ 1. 1. 0. 1.]] >>> a[1,1] = 5 >>> b[1,1] = 5 >>> print c [[ 1. 1. 1. 0.] [ 1. 1. 0. 1.]]
通过上面能够知道,这里进行是深拷贝。而不是引用指向同一位置的浅拷贝。
深拷贝数组
数组对象自带了浅拷贝和深拷贝的方法,可是一般用深拷贝多一些:
>>> a = np.ones((2,2)) >>> b = a >>> b is a True >>> c = a.copy() #深拷贝 >>> c is a False
主要的矩阵运算
转置:
>>> a = np.array([[1,0],[2,3]]) >>> print a [[1 0] [2 3]] >>> print a.transpose() [[1 2] [0 3]]
numpy.linalg模块中有非常多关于矩阵运算的方法:
特征值、特征向量:
>>> a = np.array([[1,0],[2,3]])
>>> nplg.eig(a)
(array([ 3., 1.]), array([[ 0. , 0.70710678],
[ 1. , -0.70710678]]))
原文:http://www.cnblogs.com/wzjhoutai/p/7142558.html