Go to Qiita Advent Calendar 2024 Top
LoginSignup
1
0

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?

More than 3 years have passed since last update.

@t0m0_sun(Tomo)

Numpyの基本操作をJupyter Labで書いてみた。

Last updated at Posted at 2020-04-14

この記事はかめ（＠usdatascientist）さんのブログ（https://datawokagaku.com/python_for_ds_summary/ ） に書かれているNumpyの基本操作を実際にJupyter Labを用いてコーディングしてみた、という記事です。

NumPyの基本操作まとめ

Array

import numpy as np #numpyをnpとしてインポート

np.array([1,2,3,4]) #arrayのリスト

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

python_list = (1,2,3,4) #Pythonのリスト

python_list

(1, 2, 3, 4)

np.array([[1,2,3], [4,5,6],[7,8,9]]) #arrayの入れ子のリスト

array([[1, 2, 3],
       [4, 5, 6],
       [7, 8, 9]])

[[1,2,3], [4,5,6],[7,8,9]] #Pythonの入れ子のリスト

[[1, 2, 3], [4, 5, 6], [7, 8, 9]]

Arrayの計算

arr1 = np.array([[1,2,3], [4,5,6],[7,8,9]])
arr2 = np.array([[1,2,3], [4,5,6],[7,8,9]])

arr1

array([[1, 2, 3],
       [4, 5, 6],
       [7, 8, 9]])

arr2

array([[1, 2, 3],
       [4, 5, 6],
       [7, 8, 9]])

arr1 + arr2

array([[ 2,  4,  6],
       [ 8, 10, 12],
       [14, 16, 18]])

arr1 - arr2

array([[0, 0, 0],
       [0, 0, 0],
       [0, 0, 0]])

arr1 * arr2

array([[ 1,  4,  9],
       [16, 25, 36],
       [49, 64, 81]])

arr1 / arr2

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

#arrayの足りない要素は[1,2,3]で自動的に補完される。(Broadcasting)
arr3= np.array([[1,2,3]])
arr4 = np.array([[1,2,3], [4,5,6],[7,8,9]])

arr3 + arr4 

array([[ 2,  4,  6],
       [ 5,  7,  9],
       [ 8, 10, 12]])

arr3 - arr4

array([[ 0,  0,  0],
       [-3, -3, -3],
       [-6, -6, -6]])

arr = np.array([[1,2,3], [4,5,6],[7,8,9]])

#arrayの行列のサイズを調べる
arr.shape 

(3, 3)

Indexing

indexは0から始まり、0-1番目の値まで取ってくることができる。

ndarray = np.array([[1,2], [3,4],[5,6]])
ndarray

array([[1, 2],
       [3, 4],
       [5, 6]])

ndarray[1][0]

3

ndarray[1,0]

3

Slicing

slicingは[N:M]のとき「N以上M未満」の要素を取ってくる。

arr = np.array([[1,2,3,4], [2,4,6,4], [3,5,7,4], [3,5,7,4]])

arr

array([[1, 2, 3, 4],
       [2, 4, 6, 4],
       [3, 5, 7, 4],
       [3, 5, 7, 4]])

arr[0:4, 2:4]

array([[3, 4],
       [6, 4],
       [7, 4],
       [7, 4]])

arr[:,2:] #全ての行の２列以降の要素を全て取ってくる。

array([[3, 4],
       [6, 4],
       [7, 4],
       [7, 4]])

arr = np.array([1,2,3,4,2,4,6,4,5,7,4,5,7,4])

arr[1:6:2] #2つ飛びでslicing

array([2, 4, 4])

様々なArray

np.arange[(start,)stop(,step)]

np.arange(0,5,2) #[start, stop, step] #0~5を2で区切った値のリスト

array([0, 2, 4])

np.arange(5) # start stepは省略可能。省略した場合デフォルトでそれぞれ０と１が入る。

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

np.arange(0,5,0.1) #stepは0.1刻みも可能

array([0. , 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1. , 1.1, 1.2,
       1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 2. , 2.1, 2.2, 2.3, 2.4, 2.5,
       2.6, 2.7, 2.8, 2.9, 3. , 3.1, 3.2, 3.3, 3.4, 3.5, 3.6, 3.7, 3.8,
       3.9, 4. , 4.1, 4.2, 4.3, 4.4, 4.5, 4.6, 4.7, 4.8, 4.9])

np.linspace(start, stop, num=50)

np.linspace(1, 2, 20) #1~2を２０等分する

array([1.        , 1.05263158, 1.10526316, 1.15789474, 1.21052632,
       1.26315789, 1.31578947, 1.36842105, 1.42105263, 1.47368421,
       1.52631579, 1.57894737, 1.63157895, 1.68421053, 1.73684211,
       1.78947368, 1.84210526, 1.89473684, 1.94736842, 2.        ])

.copy()

arr_copy = arr.copy() 

arr_copy

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

ndarray = np.arange(0, 5)
ndarray_copy = ndarray.copy()
print('original array is {}'.format(id(arr)))
print('copied array is {}'.format(id(arr))) #copyしたものはコピー元とは別のオブジェクト

original array is 140571396284208
copied array is 140571396284208

ndarray[:] = 100
print('original array:\n', ndarray)    #copyしないと元のarrayが更新される
print('copied array:\n', ndarray_copy) #copyすると元のarrayが更新されない

original array:
 [100 100 100 100 100]
copied array:
 [0 1 2 3 4]

def add_world(hello):
    hello += ' world'
    return hello

h_str = 'hello'
h_list = ['h', 'e', 'l', 'l', 'o']
output_str = add_world(h_str)
output_list = add_world(h_list)

print('output_str: {}'.format(output_str)) #Stringは値渡しなので直接変更する
print('output_list: {}'.format(output_list)) #listは参照渡しなので直接変更しない

print('h_str: {}'.format(h_str))
print('h_list: {}'.format(h_list))

output_str: hello world
output_list: ['h', 'e', 'l', 'l', 'o', ' ', 'w', 'o', 'r', 'l', 'd']
h_str: hello
h_list: ['h', 'e', 'l', 'l', 'o', ' ', 'w', 'o', 'r', 'l', 'd']

def change_hundred(array):
    
    array[0] = 100　#copyしないと元のarrayが更新される
    return array

def change_hundred_copy(array):
    
    array_copy = array.copy() #copyすると元のarrayが更新されない
    array_copy[0] = 100
    return array_copy

array_1 = np.arange(0, 4)
array_2 = np.arange(0,4)

output_array = change_hundred(array_1)
output_array_copy = change_hundred_copy(array_2)

print('original_array_1/n', array_1)
print('original_array2/n', array_2)

original_array_1/n [100   1   2   3]
original_array2/n [0 1 2 3]

np.zeros(shape)

shape = (2,3,5) #零行列
zeros = np.zeros(shape)
print(zeros)

[[[0. 0. 0. 0. 0.]
  [0. 0. 0. 0. 0.]
  [0. 0. 0. 0. 0.]]

 [[0. 0. 0. 0. 0.]
  [0. 0. 0. 0. 0.]
  [0. 0. 0. 0. 0.]]]

np.ones(shape)

shape = (2,3,5)
ones_array_1 = np.ones(shape)
ones_array_2 = np.ones(4)

print(ones_array_1)
print(ones_array_2)

[[[1. 1. 1. 1. 1.]
  [1. 1. 1. 1. 1.]
  [1. 1. 1. 1. 1.]]

 [[1. 1. 1. 1. 1.]
  [1. 1. 1. 1. 1.]
  [1. 1. 1. 1. 1.]]]
[1. 1. 1. 1.]

np.eye(N)

np.eye(3) #N * Nの単位行列。対角成分が全て１となる。

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

np.random.rand()

random_float = np.random.rand() # 0~1からランダムな数を返す。
random_1d = np.random.rand(3)
random_2d = np.random.rand(3, 4)

print('random_float:{}'.format(random_float))
print('random_1d:{}'.format(random_1d))
print('random_2d:{}'.format(random_2d))

random_float:0.48703321814513356/n
random_1d:[0.27824053 0.60682459 0.22001138]
random_2d:[[0.52285782 0.87272074 0.03123286 0.7921472 ]
 [0.80209903 0.26478273 0.91139303 0.63077037]
 [0.72151924 0.91234378 0.69355857 0.9969298 ]]

np.random.randn()

random_float = np.random.randn() #標準正規分布０〜１から値が返される。
random_1d = np.random.randn(3)
random_2d = np.random.randn(3, 4)

print('random_float:{}'.format(random_float))
print('random_1d:{}'.format(random_1d))
print('random_2d:{}'.format(random_2d))

random_float:-0.013248078417389888
random_1d:[-1.11606544 -0.24693187  0.17212059]
random_2d:[[-0.85896581  0.53993487  0.09458454  1.3505381 ]
 [-1.83510873  0.09966918 -0.22293729  0.58218476]
 [ 0.37403933  0.83349568  0.73407002 -0.32979339]]

np.random.randint(low[,high][,size])

np.random.randint(10, 50, size=(2,4,3)) #low以上high未満のintegerからランダムに、指定したsizeのndarrayを返す。 今回は２次元の４行３列。

array([[[30, 14, 11],
        [33, 24, 48],
        [15, 43, 41],
        [18, 26, 49]],

       [[18, 22, 16],
        [15, 10, 43],
        [43, 48, 10],
        [42, 29, 10]]])

.reshape(shape)

array = np.arange(0, 10)
print('array:{}'.format(array))
new_shape = (2, 5)
reshaped_array = array.reshape(new_shape)

print('reshaped array:{}'.format(reshaped_array))
print("reshaped array's is {}".format(reshaped_array.reshape))
print('original array is NOT changed:{}'.format(array))

array:[0 1 2 3 4 5 6 7 8 9]
reshaped array:[[0 1 2 3 4]
 [5 6 7 8 9]]
reshaped array's is <built-in method reshape of numpy.ndarray object at 0x7fd9420d8030>
original array is NOT changed:[0 1 2 3 4 5 6 7 8 9]

要素の統計量を求める

normal_dist_mat = np.random.randn(5,5)
print(normal_dist_mat) 

[[-1.51861663 -2.09873943 -0.99761607  0.95395101 -0.04577882]
 [-0.81944941  0.54339984  0.45265577 -2.56369775 -1.8300719 ]
 [ 0.63372482 -0.35763135  0.31683655  1.44185989 -1.2110421 ]
 [-1.56200024  1.17061544  1.35721624 -0.46023814  0.3496441 ]
 [ 1.09102475 -0.47551934 -0.75747612  0.34564251  1.62400795]]

.max()　最大値

print('max is')
print(normal_dist_mat.max()) 

max is
1.6240079549859363

.argmax()　最大値のindexを求める。

print('argmax is')
print(normal_dist_mat.argmax()) 

argmax is
24

.flatten()[]　index=9に格納された値を一列にする。

normal_dist_mat.flatten()[9] 

-1.83007190063398

.min()　最小値

print('min is')
print(normal_dist_mat.min()) 

min is
-2.5636977453276137

.argmin()　最小値のindexを求める。

print('argmin is')
print(normal_dist_mat.argmin()) 

argmax is
8

.mean()　平均値

print('mean is') 
print(normal_dist_mat.mean())

mean is
-0.1766919366579146

np.median(ndarray)　中央値

print('median is')
print(np.median(normal_dist_mat)) 

median is
-0.04577882007895756

.std()　標準偏差

print('standard deviation is')
print(normal_dist_mat.std())

standard deviation is
1.1634432893009636

print(normal_dist_mat.std())  #上記の関数は全てnp.関数名(ndarray)で呼ぶことも可能

1.1634432893009636

print(normal_dist_mat) #特定の行，列での統計量を求めたい場合は引数axisを指定。
print('axis=0 > {}'.format(normal_dist_mat.max(axis=0))) #axis=0を指定すると各列の統計量，
print('axis=1 > {}'.format(normal_dist_mat.max(axis=1))) #axis=1を指定すると各行の統計量を返す。

[[-1.51861663 -2.09873943 -0.99761607  0.95395101 -0.04577882]
 [-0.81944941  0.54339984  0.45265577 -2.56369775 -1.8300719 ]
 [ 0.63372482 -0.35763135  0.31683655  1.44185989 -1.2110421 ]
 [-1.56200024  1.17061544  1.35721624 -0.46023814  0.3496441 ]
 [ 1.09102475 -0.47551934 -0.75747612  0.34564251  1.62400795]]
axis=0 > [1.09102475 1.17061544 1.35721624 1.44185989 1.62400795]
axis=1 > [0.95395101 0.54339984 1.44185989 1.35721624 1.62400795]

数学で使う関数

np.exp(ndarray)　

ndarray = np.linspace(-3,3,10)
expndarray = np.exp(ndarray)
print(ndarray)
print(expndarray)

[-3.         -2.33333333 -1.66666667 -1.         -0.33333333  0.33333333
  1.          1.66666667  2.33333333  3.        ]
[ 0.04978707  0.09697197  0.1888756   0.36787944  0.71653131  1.39561243
  2.71828183  5.29449005 10.3122585  20.08553692]

np.log(nd.array)

ndarray = np.linspace(-3,3,10)
logndarray = np.log(ndarray)
print(ndarray)
print(logndarray) #負の値は対数を取れないのでnan。

[-3.         -2.33333333 -1.66666667 -1.         -0.33333333  0.33333333
  1.          1.66666667  2.33333333  3.        ]
[        nan         nan         nan         nan         nan -1.09861229
  0.          0.51082562  0.84729786  1.09861229]


/opt/anaconda3/lib/python3.7/site-packages/ipykernel_launcher.py:2: RuntimeWarning: invalid value encountered in log

print(logndarray[0])
print('nan ==None? {}'.format(logndarray[0] is None))

nan
nan ==None? False

np.isnan(logndarray[0])

True

np.e

print(np.e)
print(np.log(np.e))

2.718281828459045
1.0

ndarrayのshape操作

ndarray = np.array([[1,2,3], [4,5,6], [7,8,9]])
print('array is :{}'.format(ndarray))
print("ndarray's shape is :{}".format(ndarray.shape)) #次元の数をrankという。

array is :[[1 2 3]
 [4 5 6]
 [7 8 9]]
ndarray's shape is :(3, 3)

np.expand_dims(nd.array, axis)

expanded_ndarray = np.expand_dims(ndarray, axis=0) #次元の追加
expanded_ndarray.shape

(1, 3, 3)

np.squeeze(ndarray)

squeezed_ndarray = np.squeeze(expanded_ndarray) #shapeの次元を１つなくす。
squeezed_ndarray.shape

(3, 3)

.flatten()

flatten_array = ndarray.flatten() #1列にする
print('flatten_array is :{}'.format(flatten_array))
print('ndarray is :{}'.format(ndarray))

flatten_array is :[1 2 3 4 5 6 7 8 9]
ndarray is :[[1 2 3]
 [4 5 6]
 [7 8 9]]

Numpy Arrayを保存、読み込む

np.save('ファイルパス', ndarray)

ndarray = np.array([
    [1,2,3,4],
    [10,20,30,40],
    [100,200,300,400]
])
np.save('saved_numpy', ndarray)

np.load('ファイルパス')

loaded_numpy = np.load('saved_numpy.npy')
loaded_numpy

array([[  1,   2,   3,   4],
       [ 10,  20,  30,  40],
       [100, 200, 300, 400]])
1
0
0

Register as a new user and use Qiita more conveniently

  1. You get articles that match your needs
  2. You can efficiently read back useful information
  3. You can use dark theme
What you can do with signing up
Sign upLogin
1
0

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?