Python中NumPy簡介及使用舉例
NumPy是Python語言的一個(gè)擴(kuò)展包。支持多維數(shù)組與矩陣運(yùn)算,此外也針對數(shù)組運(yùn)算提供大量的數(shù)學(xué)函數(shù)庫。NumPy提供了與Matlab相似的功能與操作方式,因?yàn)閮烧呓詾橹弊g語言。
本文引用地址:http://www.ex-cimer.com/article/201807/383808.htmNumPy通常與SciPy(Scientific Python)和Matplotlib(繪圖庫)一起使用,這種組合廣泛用于替代Matlab,是一個(gè)流行的技術(shù)平臺(tái)。
NumPy中定義的最重要的對象是稱為ndarray的N維數(shù)組類型。它描述相同類型的元素集合,可以使用基于零的索引訪問集合中元素。基本的ndarray是使用NumPy中的數(shù)組函數(shù)創(chuàng)建的: numpy.array。
NumPy支持比Python更多種類的數(shù)值類型。NumPy數(shù)值是dtype(數(shù)據(jù)類型)對象的實(shí)例,每個(gè)對象具有唯一的特征。
以下對ndarray的介紹來自于 https://docs.scipy.org/doc/numpy/reference/arrays.ndarray.html (此鏈接可查看ndarray中包含的各種函數(shù)介紹):
An ndarray is a (usually fixed-size) multidimensional container of items of the same type and size. The number of dimensions and items in an array is defined by its shape, which is a tuple of N positive integers that specify the sizes of each dimension. The type of items in the array is specified by a separate data-type object (dtype), one of which is associated with each ndarray.
As with other container objects in Python, the contents of an ndarray can be accessed and modified by indexing or slicing the array, and via the methods and attributes of the ndarray.
Different ndarrays can share the same data, so that changes made in one ndarray may be visible in another. That is, an ndarray can be a “view” to another ndarray, and the data it is referring to is taken care of by the “base” ndarray. ndarrays can also be views to memory owned by Python strings or objects implementing the buffer or array interfaces.
以下是NumPy簡單使用例子(參考: https://wizardforcel.gitbooks.io/ts-numpy-tut/content/ ):
import numpy as np
from matplotlib import pyplot as plt
# 一維
a = np.array([1, 2, 3]); print(a) # [1 2 3]
# 等間隔數(shù)字的數(shù)組
b = np.arange(10); print(b) # [0 1 2 3 4 5 6 7 8 9]
# 二維
c = np.array([[1, 2], [3, 4]]); print(c) # [[1 2]
# [3 4]]
# ndmin指定返回?cái)?shù)組的最小維數(shù)
d = np.array([1, 2, 3, 4, 5]); print(d) # [1 2 3 4 5]
e = np.array([1, 2, 3, 4, 5], ndmin=2); print(e) # [[1 2 3 4 5]]
# dtype:數(shù)組的所需數(shù)據(jù)類型
f = np.array([1, 2, 3], dtype=complex); print(f) # [1.+0.j 2.+0.j 3.+0.j]
# 使用數(shù)組標(biāo)量類型
dt = np.dtype(np.int32); print(dt) # int32
# int8,int16,int32,int64可替換為等價(jià)的字符串'i1', 'i2', 'i4', 'i8'
dt = np.dtype('i8'); print(dt) # int64
# 調(diào)整數(shù)組shape
a = np.array([[1, 2, 3], [4, 5, 6]]); print(a); # [[1 2 3]
# [4 5 6]]
a.shape = (3, 2); print(a) # [[1 2]
# [3 4]
# [5 6]]
a = np.array([[1, 2, 3], [4, 5, 6]]); b = a.reshape(3, 2); print(b) # [[1 2]
# [3 4]
# [5 6]]
# ndim:返回?cái)?shù)組的維數(shù)
a = np.arange(24); print(a.ndim) # 1
# numpy.reshape: 在不改變數(shù)據(jù)的條件下修改形狀
b = a.reshape(2, 4, 3); print(b.ndim) # 3
# itemsize:返回?cái)?shù)組中每個(gè)元素的字節(jié)單位長度
a = np.array([1, 2, 3, 4], dtype=np.int8); print(a.itemsize) # 1
a = np.array([1, 2, 3, 4], dtype=np.float32); print(a.itemsize) # 4
# 空數(shù)組
x = np.empty([3, 2], dtype='i1'); print(x) # 數(shù)組x的元素為隨機(jī)值,因?yàn)樗鼈兾闯跏蓟?/p>
# 含有5個(gè)0的數(shù)組,若不指定類型,則默認(rèn)為float
x = np.zeros(5, dtype=np.int); print(x) # [0 0 0 0 0]
# 含有6個(gè)1的二維數(shù)組,若不指定類型,則默認(rèn)為float
x = np.ones([2, 3], dtype=int); print(x) # [[1 1 1]
# [1 1 1]]
# 將列表轉(zhuǎn)換為ndarray
x = [1, 2, 3]
a = np.asarray(x, dtype=float); print(a) # [1. 2. 3.]
# 將元組轉(zhuǎn)換為ndarray
x = (1, 2, 3)
a = np.asarray(x, dtype=complex); print(a) # [1.+0.j 2.+0.j 3.+0.j]
# 使用內(nèi)置的range()函數(shù)創(chuàng)建列表對象
x = range(5); print(x) # range(0, 5)
# 從列表中獲得迭代器
it = iter(x); print(it) #
# 使用迭代器創(chuàng)建ndarray, fromiter函數(shù)從任何可迭代對象構(gòu)建一個(gè)ndarray對象,返回一個(gè)新的一維數(shù)組
y = np.fromiter(it, dtype=float); print(y) # [0. 1. 2. 3. 4.]
# arange函數(shù)返回ndarray對象,包含給定范圍內(nèi)的等間隔值
# numpy.arange(start, stop, step, dtype), start起始值,默認(rèn)為0;stop終止值,不包含; step間隔,默認(rèn)為1
x = np.arange(4, dtype=float); print(x) # [0. 1. 2. 3.]
x = np.arange(10, 20, 2); print(x) # [10 12 14 16 18]
# numpy.linspace,此函數(shù)類似于arange,在此函數(shù)中,指定了范圍之間的均勻間隔數(shù)量,而不是步長
# numpy.linspace(start, stop, num, endpoint, retstep, dtype)
# start,起始值;stop,終止值,如果endpoint為true,該值包含于序列中;num,要生成的等間隔樣例數(shù)量,默認(rèn)為50;
# endpoint,序列中是否包含stop值,默認(rèn)為true;retstep,如果為true,返回樣例,以及連續(xù)數(shù)字之間的步長
x = np.linspace(10, 20, 5); print(x) # [10. 12.5 15. 17.5 20.]
x = np.linspace(10, 20, 5, endpoint=False); print(x) # [10. 12. 14. 16. 18.]
x = np.linspace(1,2,5, retstep=True); print(x) # (array([ 1. , 1.25, 1.5 , 1.75, 2. ]), 0.25)
# numpy.logspace,此函數(shù)返回ndarray對象,包含在對數(shù)刻度上均勻分布的數(shù)字.刻度的開始和結(jié)束端點(diǎn)是某個(gè)底數(shù)的冪,通常為10
# numpy.logscale(start, stop, num, endpoint, base, dtype)
# base,對數(shù)空間的底數(shù),默認(rèn)為10;其它參數(shù)同numpy.linspace
a = np.logspace(1.0, 2.0, num=5); print(a) # [10. 17.7827941 31.6227766 56.23413252 100.]
a = np.logspace(1, 10, num=5, base=2); print(a) # [2. 9.51365692 45.254834 215.2694823 1024.]
# ndarray對象的內(nèi)容可以通過索引或切片來訪問和修改,就像Python的內(nèi)置容器對象一樣;
# 基本切片:通過將start、stop和step參數(shù)提供給內(nèi)置的slice函數(shù)來構(gòu)造一個(gè)Python slice對象,用來提前數(shù)組的一部分
a = np.arange(10); s = slice(2,7,2); print(a[s]) # [2 4 6]
# 通過將由冒號(hào)分隔的切片參數(shù)(start:stop:step)直接提供給ndarray對象,也可以獲得相同的結(jié)果
a = np.arange(10); b = a[2:7:2]; print(b) # [2 4 6]
a = np.arange(10); b = a[2:]; print(b) # [2 3 4 5 6 7 8 9]
a = np.arange(10); b = a[2:5]; print(b) # [2 3 4]
# 切片還可以包括省略號(hào)(...),來使選擇元組的長度與數(shù)組的維度相同
a = np.array([[1,2,3],[3,4,5],[4,5,6]])
b = a[..., 1]; print(b) # [2 4 5]
c = a[1, ...]; print(c) # [3 4 5]
# 高級(jí)索引:如果一個(gè)ndarray是非元組序列,數(shù)據(jù)類型為整數(shù)或布爾值的ndarray,或者至少一個(gè)元素為
# 序列對象的元組,我們就能夠用它來索引ndarray,高級(jí)索引始終返回?cái)?shù)據(jù)的副本
# 高級(jí)索引:整數(shù):基于N維索引來獲取數(shù)組中任意元素
x = np.array([[1, 2], [3, 4], [5, 6]])
# y中包括數(shù)組x中(0,0), (1,1), (2,0)位置處的元素
y = x[[0,1,2], [0,1,0]]; print(y) # [1 4 5]
# 高級(jí)索引:布爾值:當(dāng)結(jié)果對象是布爾運(yùn)算的結(jié)果時(shí),將使用此類型的高級(jí)索引
x = np.array([[0, 1, 2], [3, 4, 5], [6, 7, 8], [9, 10, 11]])
y = x[x > 5]; print(y) # [6 7 8 9 10 11]
# ~(取補(bǔ)運(yùn)算符)來過濾NaN
x = np.array([np.nan, 1, 2, np.nan, 3, 4, 5])
y = x[~np.isnan(x)]; print(y) # [1. 2. 3. 4. 5.]
# 從數(shù)組中過濾掉非復(fù)數(shù)元素
x = np.array([1, 2+6j, 5, 3.5+5j])
y = x[np.iscomplex(x)]; print(y) # [2.0+6.j 3.5+5.j]
# 廣播:是指NumPy在算術(shù)運(yùn)算期間處理不同形狀的數(shù)組的能力, 對數(shù)組的算術(shù)運(yùn)算通常在相應(yīng)的元素上運(yùn)行
# 如果兩個(gè)數(shù)組的維數(shù)不相同,則元素到元素的操作是不可能的。然而,在NumPy中仍然可以對形狀不相似的數(shù)組進(jìn)行操作,因?yàn)樗鼡碛袕V播功能。
# 較小的數(shù)組會(huì)廣播到較大數(shù)組的大小,以便使它們的形狀可兼容
a = np.array([1, 2, 3, 4])
b = np.array([10, 20, 30, 40])
c = a * b; print(c) # [10 40 90 160]
a = np.array([[0.0,0.0,0.0],[10.0,10.0,10.0]])
b = np.array([1.0,2.0,3.0])
c = a + b; print(c) # [[1.0 2.0 3.0]
# [11. 12. 13.]]
# 數(shù)組上的迭代:NumPy包包含一個(gè)迭代器對象numpy.nditer。它是一個(gè)有效的多維迭代器對象,可以用于在數(shù)組上進(jìn)行迭代。
# 數(shù)組的每個(gè)元素可使用Python的標(biāo)準(zhǔn)Iterator接口來訪問
a = np.arange(0, 60, 5)
a = a.reshape(3,4)
for x in np.nditer(a):
print(x, end=' ') # 0 5 10 15 20 25 30 35 40 45 50 55
print('n')
# 修改數(shù)組的值: nditer對象的一個(gè)可選參數(shù)op_flags,其默認(rèn)值為只讀,但可以設(shè)置為讀寫或只寫模式.這將允許使用此迭代器修改數(shù)組元素
for x in np.nditer(a, op_flags=['readwrite']):
x[...]=2*x; print(x, end=' ') # 0 10 20 30 40 50 60 70 80 90 100 110
print('n')
# numpy.raval:返回展開的一維數(shù)組,并且按需生成副本。返回的數(shù)組和輸入數(shù)組擁有相同數(shù)據(jù)類型
a = np.arange(8).reshape(2,4)
b = a.ravel(); print(b) # [0 1 2 3 4 5 6 7]
# numpy.unique: 返回輸入數(shù)組中的去重元素?cái)?shù)組
a = np.array([5, 2, 6, 2, 7, 5, 6, 8, 2, 9])
u = np.unique(a); print(u) # [2 5 6 7 8 9]
# 位操作:bitwise_and, bitwise_or, invert, left_shift, right_shift
a,b = 13,17; print(bin(a), bin(b)) # 0b1101 0b10001
c = np.bitwise_and(13, 17); print(c) # 1
c = np.bitwise_or(13, 17); print(c) # 29
# 字符串函數(shù):add, multiply, center, capitalize, title, lower, upper, split, splitlines, strip, join, replace, decode, encode
print(np.char.add(['hello'],[' Spring'])) # ['hell Spring']
print(np.char.multiply('Hello ',3)) # Hello Hello Hello
# numpy.char.center: 此函數(shù)返回所需寬度的數(shù)組,以便輸入字符串位于中心,并使用fillchar在左側(cè)和右側(cè)進(jìn)行填充
print(np.char.center('hello', 20, fillchar = '*')) # *******hello********
a = np.char.encode('hello', 'cp500'); print(a) # b'x88x85x93x93x96'
b = np.char.decode(a, 'cp500'); print(b) # hello
# 三角函數(shù):sin, cos, tan, arcsin, arccos, arctan
a = np.array([0, 30, 45, 60, 90])
b = np.sin(a*np.pi/180); print(b) # [ 0. 0.5 0.70710678 0.8660254 1.]
# 舍入函數(shù): around, floor, ceil
a = np.array([1.0, 5.55, 123, 0.567, 25.532])
b = np.around(a); print(b) # [1. 6. 123. 1. 26.]
# 算數(shù)運(yùn)算:add, subtract, multiply, divide, reciprocal, power, mod 輸入數(shù)組必須具有相同的形狀或符合數(shù)組廣播規(guī)則
a, b = [5, 6], [7, 10]
c = np.subtract(a, b); print(c) # [-2 -4]
# 統(tǒng)計(jì)函數(shù):用于從數(shù)組中給定的元素中查找最小,最大,百分標(biāo)準(zhǔn)差和方差等, amin, amax, ptp, percentile, median, mean, average, std
a = np.array([1, 2, 3, 4, 5])
print(np.amin(a)) # 1
print(np.median(a)) # 3.0
print(np.mean(a)) # 3.0
# 副本和視圖: 在執(zhí)行函數(shù)時(shí),其中一些返回輸入數(shù)組的副本,而另一些返回視圖。 當(dāng)內(nèi)容物理存儲(chǔ)在另一個(gè)位置時(shí),稱為副本。
# 另一方面,如果提供了相同內(nèi)存內(nèi)容的不同視圖,我們將其稱為視圖
a = np.arange(6); print(a) # [0 1 2 3 4 5]
print(id(a)) # 54667664
b = a
print(id(b)) # 54667664
b.shape = 2,3
print(a); # [[0 1 2]
# [3 4 5]]
# IO: ndarray對象可以保存到磁盤文件并從磁盤文件加載
# load()和save()函數(shù)處理NumPy二進(jìn)制文件(帶npy擴(kuò)展名)
# loadtxt()和savetxt()函數(shù)處理正常的文本文件
a = np.array([1, 2, 3, 4, 5])
np.save('E:/GitCode/Python_Test/test_data/outfile.npy', a)
b = np.load('E:/GitCode/Python_Test/test_data/outfile.npy')
print(b) # [1 2 3 4 5]
np.savetxt('E:/GitCode/Python_Test/test_data/outfile.txt', a)
b = np.loadtxt('E:/GitCode/Python_Test/test_data/outfile.txt')
print(b) # [1. 2. 3. 4. 5.]
# Matplotlib是Python的繪圖庫,在http://matplotlib.org/examples/ 中含有大量的matplotlib使用用例
x = np.arange(1,11)
y = 2 * x + 5
plt.title(Matplotlib demo)
plt.xlabel(x axis caption)
plt.ylabel(y axis caption)
plt.plot(x,y, 'ob')
plt.show()
評論