1. Boolean indexes
1
2
3
4
5
6
7
8
9
10
11
12
13
import numpy as np
a = np.arange(12).reshape(3,4)
print('-' * 20)
print(a)
print('-' * 20)
# 数组a中所有大于5的位置为True, 其余为False
print(a>5)
print('-' * 20)
# 将数组a中所有大于5的位置赋值为999
a[a>5]=999
print(a)
--------------------
[[ 0 1 2 3]
[ 4 5 6 7]
[ 8 9 10 11]]
-------a > 5-------
[[False False False False]
[False False True True]
[ True True True True]]
-------a[a>5]=999-------
[[ 0 1 2 3]
[ 4 5 999 999]
[999 999 999 999]]
2. 三元运算符
np.where
1
2
3
4
5
6
7
8
9
import numpy as np
a = np.arange(12).reshape(3,4)
print('-' * 20)
print(a)
print('-' * 20)
# a中大于5的位置为999, 其余为111
np.where(a>5, 999, 111)
--------------------
[[ 0 1 2 3]
[ 4 5 6 7]
[ 8 9 10 11]]
-------np.where-------
array([[111, 111, 111, 111],
[111, 111, 999, 999],
[999, 999, 999, 999]])
3. clip
1
2
3
4
5
6
7
8
9
import numpy as np
a = np.arange(12).reshape(3,4)
print('-' * 20)
print(a)
print('-' * 20)
# a小于4的全都赋值为4，大于8的全都赋值为8
a.clip(4, 8)
--------------------
[[ 0 1 2 3]
[ 4 5 6 7]
[ 8 9 10 11]]
-------clip-------
array([[4, 4, 4, 4],
[4, 5, 6, 7],
[8, 8, 8, 8]])
4. stack
np.vstack() 列拼接 and np.hstack() 行拼接
1
2
3
4
5
6
7
8
9
10
11
12
import numpy as np
t1 = np.arange(12).reshape(3,4)
t2 = np.arange(12).reshape(3,4)
print('-' * 20)
print(t1)
print('-' * 20)
print(t2)
print('-' * 20)
np.vstack((t1,t2))
np.hstack((t1,t2))
--------------------
[[ 0 1 2 3]
[ 4 5 6 7]
[ 8 9 10 11]]
--------------------
[[ 0 1 2 3]
[ 4 5 6 7]
[ 8 9 10 11]]
-------列拼接-------
array([[ 0, 1, 2, 3],
[ 4, 5, 6, 7],
[ 8, 9, 10, 11],
[ 0, 1, 2, 3],
[ 4, 5, 6, 7],
[ 8, 9, 10, 11]])
-------行拼接-------
array([[ 0, 1, 2, 3, 0, 1, 2, 3],
[ 4, 5, 6, 7, 4, 5, 6, 7],
[ 8, 9, 10, 11, 8, 9, 10, 11]])
5. 行列交换
1
2
3
4
5
6
7
8
9
10
11
12
import numpy as np
t1 = np.arange(12).reshape(3,4)
print('-' * 20)
print(t1)
print('-' * 20)
t1[[1,2],:] = t1[[2,1],:] # 行交换
print(t1)
print('-' * 20)
t1[:,[0,2]] = t1[:,[2,0]] # 交换
print(t1)
-------原始t1-------
[[ 0 1 2 3]
[ 4 5 6 7]
[ 8 9 10 11]]
-------行交换-------
[[ 0 1 2 3]
[ 8 9 10 11]
[ 4 5 6 7]]
-------列交换-------
[[ 2 1 0 3]
[10 9 8 11]
[ 6 5 4 7]]
6. squeeze and unsqueeze
squeeze作用: 从数组的形状中删除单维度条目，即把shape中为1的维度去掉
1
2
3
4
5
import numpy as np
a = np.arange(10).reshape(1,10)
print(a)
print(a.shape)
array([[0, 1, 2, 3, 4, 5, 6, 7, 8, 9]])
(1, 10)
1
2
3
b = np.squeeze(a)
print(b)
print(b.shape)
array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
(10,)
对维度不为1的维度没有用
unsqueeze作用: 增加一个维度
例如: a 的维度为(3, 4), a.unsqueeze(1) 就是在第二维增加一个维度使其变为(3, 1, 4)
7. numpy的copy和view
a = b 完全不复制, a和b相互影响
1
2
3
4
5
6
7
8
9
10
11
12
13
import numpy as np
a = np.arange(12).reshape(3,4)
print('-' * 20)
print(a)
print('-' * 20)
b = a
b[b>5] = 999
print(a)
print('-' * 20)
a[a<2] = 111
print(b)
-------初始的a-------
[[ 0 1 2 3]
[ 4 5 6 7]
[ 8 9 10 11]]
-------改变b看a的值-------
[[ 0 1 2 3]
[ 4 5 999 999]
[999 999 999 999]]
-------改变a看b的值-------
[[111 111 2 3]
[ 4 5 999 999]
[999 999 999 999]]
a=b[:], 视图的操作，一种切片，会创建新的对象a, 但是a的数据完全由b保管, 他们两个的数据变化是一致的，和上个操作a=b一样，都是相互影响
a = b.copy()复制，a和b互不影响
8. torch.sum
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
import torch
a = torch.tensor([[[[1,2,3],[4,5,6]],[[7,8,9],[10,11,12]]], [[[1,2,3],[4,5,6]],[[7,8,9],[10,11,12]]]])
b = torch.tensor([[[[1,0,0],[0,1,0]],[[0,0,1],[0,1,0]]], [[[1,0,0],[0,1,0]],[[0,0,1],[0,1,0]]]])
print(a)
print('-'*20)
print(a.shape)
print('-'*20)
print(torch.sum(a, 0))
print('-'*20)
print(torch.sum(a, 1))
print('-'*20)
print(torch.sum(a, [0,1]))
print('='*20)
print(a*b)
print('='*30)
print(torch.sum(a*b, 0))
print(torch.sum(a*b, [0,1]))
tensor([[[[ 1, 2, 3],
[ 4, 5, 6]],
[[ 7, 8, 9],
[10, 11, 12]]],
[[[ 1, 2, 3],
[ 4, 5, 6]],
[[ 7, 8, 9],
[10, 11, 12]]]])
--------------------
torch.Size([2, 2, 2, 3])
--------------------
tensor([[[ 2, 4, 6],
[ 8, 10, 12]],
[[14, 16, 18],
[20, 22, 24]]])
--------------------
tensor([[[ 8, 10, 12],
[14, 16, 18]],
[[ 8, 10, 12],
[14, 16, 18]]])
--------------------
tensor([[16, 20, 24],
[28, 32, 36]])
====================
tensor([[[[ 1, 0, 0],
[ 0, 5, 0]],
[[ 0, 0, 9],
[ 0, 11, 0]]],
[[[ 1, 0, 0],
[ 0, 5, 0]],
[[ 0, 0, 9],
[ 0, 11, 0]]]])
==============================
tensor([[[ 2, 0, 0],
[ 0, 10, 0]],
[[ 0, 0, 18],
[ 0, 22, 0]]])
tensor([[ 2, 0, 18],
[ 0, 32, 0]])
9. sundry
- np.argmax(t, axis=0) 获取数组t第0维最大值的位置
- np.argmin(t, axis=0) 获取数组t第0维最小值的位置
- np.zeros((3,4)) 创建全0数组
- np.ones((3,4)) 创建全1数组
- np.eye(3) 创建一个对角线全为1，其余全为0的正方形矩阵
- np.seed(s) 随机数种子，s是给定的种子值。因为计算机生成的是伪随机数，所以通过设定相同的随机数种子，可以每次生成相同的随机数
