Python資料分析--Numpy常用函式介紹(7)--Numpy中矩陣和通用函式

PursuitingPeak發表於2022-06-08

在NumPy中,矩陣是 ndarray 的子類,與數學概念中的矩陣一樣,NumPy中的矩陣也是二維的,可以使用 mat 、 matrix 以及 bmat 函式來建立矩陣。

一、建立矩陣

mat 函式建立矩陣時,若輸入已為 matrix 或 ndarray 物件,則不會為它們建立副本。 因此,呼叫 mat() 函式和呼叫 matrix(data, copy=False) 等價。

1) 在建立矩陣的專用字串中,矩陣的行與行之間用分號隔開,行內的元素之間用空格隔開。使用如下的字串呼叫 mat 函式建立矩陣:

import numpy as np

A = np.mat('1 2 3; 4 5 6; 7 8 9')
print("Creation from string:", A)

執行結果:

Creation from string: 
[[1 2 3]
 [4 5 6]
 [7 8 9]]

2)用T屬性獲取轉置矩陣

print("transpose A:", A.T)  # 用T屬性獲取轉置矩陣

3)用I屬性獲取逆矩陣

print("Inverse A:", A.I)  # 用I屬性獲取逆矩陣

4)用NumPy陣列進行建立矩陣

B = np.mat(np.arange(9).reshape(3, 3))
print("Creation from array:", B)#使用NumPy陣列進行建立

上述執行結果:

Creation from string: 
[[1 2 3] [4 5 6] [7 8 9]] transpose A:
[[
1 4 7] [2 5 8] [3 6 9]] Inverse A:
[[
3.15251974e+15 -6.30503948e+15 3.15251974e+15] [-6.30503948e+15 1.26100790e+16 -6.30503948e+15] [ 3.15251974e+15 -6.30503948e+15 3.15251974e+15]] Creation from array:
[[0 1 2] [3 4 5] [6 7 8]]

二、從已有矩陣建立新矩陣

希望利用一些已有的較小的矩陣來建立一個新的大矩陣。這可以用 bmat 函式來實現。這裡的 b 表示“分塊”, bmat 即分塊矩陣(block matrix)。

1)先建立一個3*3的單位矩陣:

C = np.eye(3)
print("C:",C)

執行結果:

C: 
[[1. 0. 0.]
 [0. 1. 0.]
 [0. 0. 1.]]

2)建立一個與C同型的矩陣,乘以2

D = 2 * C
print ("D:",D)

執行結果:

D: 
[[2. 0. 0.] [0. 2. 0.] [0. 0. 2.]]

3)使用字串建立複合矩陣:

字串的格式與 mat 函式中一致,只是在這裡你可以用矩陣變數名代替數字:

print("Compound matrix\n", np.bmat("C D;C D"))

執行結果:

Compound matrix:
 [[1. 0. 0. 2. 0. 0.]
 [0. 1. 0. 0. 2. 0.]
 [0. 0. 1. 0. 0. 2.]
 [1. 0. 0. 2. 0. 0.]
 [0. 1. 0. 0. 2. 0.]
 [0. 0. 1. 0. 0. 2.]]

三、通用函式

通用函式的輸入是一組標量,輸出也是一組標量,它們通常可以對應於基本數學運算,如加、減、乘、除等。

1、使用NumPy中的 frompyfunc 函式,通過一個Python函式來建立通用函式,步驟如下:

1)定義一個回答某個問題的Python函式

2)用 zeros_like 函式建立一個和 a 形狀相同,並且元素全部為0的陣列 result

3)將剛生成的陣列中的所有元素設定其值為42

2、在 add 上呼叫通用函式的方法

通用函式並非真正的函式,而是能夠表示函式的物件。通用函式有四個方法,不過這些方法只對輸入兩個引數、輸出一個引數的ufunc物件有效,例如 add 函式。

其他不符合條件的ufunc物件呼叫這些方法時將丟擲 ValueError 異常。因此只能在二元通用函式上呼叫這些方法。以下將逐一介紹這4個方法:

 reduce()、accumulate()、 reduceat()、outer()

1) 沿著指定的軸,在連續的陣列元素之間遞迴呼叫通用函式,即可得到輸入陣列的規約(reduce)計算結果。

對於 add 函式,其對陣列的reduce計算結果等價於對陣列元素求和。呼叫reduce 方法:

a = np.arange(9)
print("Reduce:", np.add.reduce(a)) #呼叫add函式的reduce方法

執行結果:

Reduce 36

2) accumulate 方法同樣可以遞迴作用於輸入陣列

在 add 函式上呼叫 accumulate 方法,等價於直接呼叫 cumsum 函式。在 add 函式上呼叫 accumulate 方法:

print( "Accumulate", np.add.accumulate(a)) #呼叫add函式的accumulate方法

執行結果:

Accumulate [ 0  1  3  6 10 15 21 28 36]

3)educeat 方法需要輸入一個陣列以及一個索引值列表作為引數。

print ("Reduceat", np.add.reduceat(a, [0, 5, 2, 7]))

educeat 方法的作用是,在數列a中,分別計算索引間的累加,比如上述的 [0, 5, 2, 7],分別計算索引0-5,5-2(5>2,所以直接取索引為5的資料),2-7,7-(-1) 等四組序列形成的

比如,0-5就是計算A-E列中的資料,結果為10;5-2,直接取索引為5,即F的資料5;2-7,即B-G的計算結果為20;7-(-1)即索引7到最後,也即H、I的計算結果為15。

4)outer 方法

返回一個陣列,它的秩(rank)等於兩個輸入陣列的秩的和。它會作用於兩個輸入陣列之間存在的所有元素對。在 add 函式上呼叫 outer 方法:

print("Outer:\n", np.add.outer(np.arange(3), a))

執行結果:

Outer:
 [[ 0  1  2  3  4  5  6  7  8]
 [ 1  2  3  4  5  6  7  8  9]
 [ 2  3  4  5  6  7  8  9 10]]

四、算術運算

在NumPy中,基本算術運算子+、-和 * 隱式關聯著通用函式 add 、 subtract 和 multiply ,對NumPy陣列使用這些算術運算子時,對應的通用函式將自動被呼叫。除法包含

的過程則較為複雜,在陣列的除法運算中涉及

三個通用函式 divide 、 true_divide 和floor_division ,以及兩個對應的運算子 / 和 // 。
1、除法運算:

import numpy as np

a = np.array([2, 6, 5])
b = np.array([1, 2, 3])

print("Divide:\n", np.divide(a, b), np.divide(b, a))

 除了divide()函式外,還有floor_divide(),以及運算子‘/’和‘//’,(‘/’和‘//’分別和divide和floor_divide作用一樣)如下程式碼:

import numpy as np

a = np.array([2, 6, 5])
b = np.array([1, 2, 3])

print("Divide:\n", np.divide(a, b), np.divide(b, a))
print("True Divide:\n", np.true_divide(a, b), np.true_divide(b, a))#回除法的浮點數結果而不作截斷

print("Floor Divide:\n", np.floor_divide(a, b), np.floor_divide(b, a))  #返回整數結果
c = 3.14*b
print("Floor Divide2:\n", np.floor_divide(c, b), np.floor_divide(b, c)) #返回整數結果

print( "/ operator:\n", a/b, b/a)  # "/"運算子相當於呼叫 divide 函式

print( "// operator:\n", a//b, b//a) #運算子//對應於floor_divide 函式
print( "// operator2:\n", c//b, b//c) 

執行結果:

Divide:
 [2.         3.         1.66666667] [0.5        0.33333333 0.6       ]
True Divide:
 [2.         3.         1.66666667] [0.5        0.33333333 0.6       ]
Floor Divide:
 [2 3 1] [0 0 0]
Floor Divide2:
 [3. 3. 3.] [0. 0. 0.]
/ operator:
 [2.         3.         1.66666667] [0.5        0.33333333 0.6       ]
// operator:
 [2 3 1] [0 0 0]
// operator2:
 [3. 3. 3.] [0. 0. 0.]

 2、模運算

計算模數或者餘數,可以使用NumPy中的 mod 、 remainder 和 fmod 函式。當然,也可以使用 % 運算子。這些函式的主要差異在於處理負數的方式。

a = np.arange(-4, 4)
print('a:',a)
print ("Remainder", np.remainder(a, 2)) # remainder 函式逐個返回兩個陣列中元素相除後的餘數
print ("Mod", np.mod(a, 2))  # mod 函式與 remainder 函式的功能完全一致
print ("% operator", a % 2)  # % 操作符僅僅是 remainder 函式的簡寫

print ("Fmod", np.fmod(a, 2))# fmod 函式處理負數的方式與 remainder 、 mod 和 % 不同

執行結果:

a: [-4 -3 -2 -1  0  1  2  3]
Remainder [0 1 0 1 0 1 0 1]
Mod [0 1 0 1 0 1 0 1]
% operator [0 1 0 1 0 1 0 1]
Fmod [ 0 -1  0 -1  0  1  0  1]

實際程式碼執行如下:

 

 

 

 

相關文章