Numpy common functions and usage
This article will be supplemented and updated in the future!
References:
1. Basic use of Numpy
2. Detailed explanation of numpy file storage .npy .npz files
3. np.hstack usage
4. Diagram
5. Rookie tutorial
6. np.argmax() of python and understanding of axis=0 or 1
1. np.delete()
删除指定行
np.delete(x, i, axis=0) #删除x矩阵 第i行
2. np.where()
返回输入数组中满足给定条件的元素的索引, the return value istuple type。
import numpy as np
x = np.arange(9.).reshape(3, 3)
print ('我们的数组是：')
print (x)
print ( '大于 3 的元素的索引：')
y = np.where(x > 3)
print (y)
print ('使用这些索引来获取满足条件的元素：')
print (x[y])
结果：
我们的数组是：
[[0. 1. 2.]
[3. 4. 5.]
[6. 7. 8.]]
大于 3 的元素的索引（元组）：
(array([1, 1, 2, 2, 2]), array([1, 2, 0, 1, 2]))
使用这些索引来获取满足条件的元素：
[4. 5. 6. 7. 8.]
3. np.load()
读取npy文件.
By default, arrays are saved in uncompressed raw binary format in files with the extension .npy. .npz is a compressed file.
Reference : numpy file storage .npy .npz file details
>>> np.save('/tmp/123', np.array([[1, 2, 3], [4, 5, 6]]))
>>> np.load('/tmp/123.npy')
array([[1, 2, 3],
[4, 5, 6]])
>>> a=np.array([[1, 2, 3], [4, 5, 6]])
>>> b=np.array([1, 2])
>>> np.savez('/tmp/123.npz', a=a, b=b)
>>> data = np.load('/tmp/123.npz')
>>> data['a']
array([[1, 2, 3],
[4, 5, 6]])
>>> data['b']
array([1, 2])
>>> data.close()
>>> X = np.load('/tmp/123.npy', mmap_mode='r')
>>> X[1, :]
memmap([4, 5, 6])
4.np.item()
可以获取在numpy数组上给定索引处找到的数据元素。
# import the important module in python
import numpy as np
# make an array with numpy
gfg = np.array([[1, 2, 3, 4, 5],
[6, 5, 4, 3, 2]])
print(gfg.item((1, 2))) ## 结果为4
5. Concatenate Arrays
| function | describe |
|---|---|
| concatenate() | concatenates sequences of arrays along existing axes |
| stack() | Add a new dimension to stack |
| hstack() | horizontal stack |
| vstack() | vertical stack |
| dstack() | deep stack |
Reference:
np.hstack Usage
Diagram
Novice Tutorial
eg1.
a = np.array([[1,2],[3,4]])
b = np.array([[5,6],[7,8]])
print ('沿轴 0 连接两个数组：')
print (np.concatenate((a,b))) # 默认是0轴
print ('沿轴 1 连接两个数组：')
print (np.concatenate((a,b),axis = 1))
#结果
第一个数组：
[[1 2]
[3 4]]
第二个数组：
[[5 6]
[7 8]]
沿轴 0 连接两个数组：
[[1 2]
[3 4]
[5 6]
[7 8]]
沿轴 1 连接两个数组：
[[1 2 5 6]
[3 4 7 8]]
eg2.
print(np.stack((arr1, arr2),0)) # 堆叠后维度增加，axis=0/1效果相同
print(np.vstack((arr1, arr2)))
print(np.hstack((arr1, arr2)))
print(np.dstack((arr1, arr2)))
# 结果
[[[1 2]
[3 4]]
[[5 6]
[7 8]]]
[[ 1 2 3]
[ 4 5 6]
[ 7 8 9]
[10 11 12]]
[[ 1 2 3 7 8 9]
[ 4 5 6 10 11 12]]
[[[ 1 7]
[ 2 8]
[ 3 9]]
[[ 4 10]
[ 5 11]
[ 6 12]]]
6. Modify the array dimension
6.1 np.squeeze()
从数组的形状中删除一维条目，及shape中是１的维度。
For example, the dimensions are (1,2,5)->(2,5), (3,4)->(3,4)
np.squeeze()
eg.
c = np.arange(10).reshape(2,5)
np.squeeze(c)
c.shape # (2,5)
d = np.arange(10).reshape(1,2,5)
np.squeeze(d)
d.shape # (2,5)
x = np.arange(9).reshape(1,3,3)
y = np.squeeze(x)
print (x.shape, y.shape)
# 结果
(1, 3, 3) (3, 3)
6.2 np.expand_dims()
Extend the shape of the array.
cat_mask = np.expand_dims(cat_mask, 0) # cat_mask从原来的[256,256]变为[1,256,256]
7. np.unique()
用于去除数组中的重复元素。
Function prototype:
numpy.unique(arr, return_index, return_inverse, return_counts)
- arr: input array, if it is not a one-dimensional array, it will be expanded
- return_index: If true, return the position (subscript) of the new list element in the old list, and store it in the form of a list
- return_inverse: If true, return the position (subscript) of the old list element in the new list, and store it in the form of a list
- return_counts: If true, returns the number of occurrences of elements in the deduplicated array in the original array
eg.
a = np.array([5,2,6,2,7,5,6,8,2,9])
print (a) # [5 2 6 2 7 5 6 8 2 9]
u = np.unique(a)
print (u) # [2 5 6 7 8 9], 顺便排序了
u,indices = np.unique(a, return_index = True)
print (indices) # [1 0 2 4 7 9]
8. Sorting, conditional filter function
用于给多个序列进行排序，想象成电子表格，优先排序后面的列。
numpy.lexsort()
eg.
nm = ('raju','anil','ravi','amar')
dv = ('f.y.', 's.y.', 's.y.', 'f.y.')
ind = np.lexsort((dv,nm))
print (ind)
print ([nm[i] + ", " + dv[i] for i in ind])
# 结果
[3 1 0 2]
['amar, f.y.', 'anil, s.y.', 'raju, f.y.', 'ravi, s.y.']
9. Contiguous memory array
In Numpy, randomly initialized arrays are C continuous by default, and the continuity will be changed after irregular slice operations. ascontiguousarray函数将一个内存不连续存储的数组转换为内存连续存储的数组, makingrun faster。
np.ascontiguousarray()
eg.
>>> x = np.arange(8).reshape(2,4)
>>> np.ascontiguousarray(x, dtype=np.float32)
array([[0., 1., 2., 3.],
[4., 5., 6., 7.]], dtype=float32)
>>> x.flags['C_CONTIGUOUS']
True
10. Statistics frequency
像直方图一样，统计元素出现频率，并返回ndarray。
np.bincount()
Reference : np.bincount usage
eg.
x=np.array([0,1,1,3,2,1,7])
print(np.bincount(x))
# 结果
[1 3 1 1 0 0 0 1]
11. np.argmax()
返回数组中较大值所在的索引，这里针对一维、二维、三维或者高维数据。
Reference: python's np.argmax() and understanding of axis=0 or 1
import numpy as np
# 三维数组
a = np.array([
[
[1, 5, 5, 2],
[9, -6, 2, 8],
[-3, 7, -9, 1]
],
[
[-1, 5, -5, 2],
[9, 6, 2, 8],
[3, 7, 9, 1]
]
])
print(np.argmax(a, axis=0))
结果：从[2,3,4]->[3,4]
[[0 0 0 0]
[0 1 0 0]
[1 0 1 0]]