np.tile 和np.newaxis
output
array([[ 0.24747071, -0.43886742],
[-0.03916734, -0.70580089],
[ 0.00462337, -0.51431584],
...,
[ 0.15071507, -0.57029653],
[ 0.06246116, -0.33766761],
[ 0.08218585, -0.59906501]], dtype=float32)
ipdb> np.shape(output)
(64, 2)
ipdb> np.max(output, axis=1)[:,np.newaxis]
array([[ 0.24747071],
[-0.03916734],
[ 0.00462337],
...,
[ 0.15071507],
[ 0.06246116],
[ 0.08218585]], dtype=float32)
ipdb> np.tile(np.max(output, axis=1)[:,np.newaxis], [1,2]))
*** SyntaxError: invalid syntax (, line 1)
ipdb> np.tile(np.max(output, axis=1)[:,np.newaxis], [1,2])
array([[ 0.24747071, 0.24747071],
[-0.03916734, -0.03916734],
[ 0.00462337, 0.00462337],
...,
[ 0.15071507, 0.15071507],
[ 0.06246116, 0.06246116],
[ 0.08218585, 0.08218585]], dtype=float32)
ipdb> output
array([[ 0.24747071, -0.43886742],
[-0.03916734, -0.70580089],
[ 0.00462337, -0.51431584],
...,
[ 0.15071507, -0.57029653],
[ 0.06246116, -0.33766761],
[ 0.08218585, -0.59906501]], dtype=float32)
ipdb> np.max(output, axis=1)
array([ 0.24747071, -0.03916734, 0.00462337, ..., 0.15071507,
0.06246116, 0.08218585], dtype=float32)
ipdb> np.exp(output - np.tile(np.max(output, axis=1)[:,np.newaxis], [1,2]))
array([[ 1. , 0.50341612],
[ 1. , 0.51343411],
[ 1. , 0.59515154],
...,
[ 1. , 0.48626012],
[ 1. , 0.67023373],
[ 1. , 0.50598365]], dtype=float32)
1- np.newaxis
np.newaxis的功能是插入新维度，看下面的例子：
print a.shape
print a
输出结果
(5,)
[1 2 3 4 5]
可以看出a是一个一维数组,
a=np.array([1,2,3,4,5])
b=a[np.newaxis,:]
print a.shape,b.shape
print a
print b
输出结果：
(5,) (1, 5)
[1 2 3 4 5]
[[1 2 3 4 5]]
x_data=np.linspace(-1,1,300)[:,np.newaxis]
a=np.array([1,2,3,4,5])
b=a[:,np.newaxis]
print a.shape,b.shape
print a
print b
输出结果
(5,) (5, 1)
[1 2 3 4 5]
[[1]
[2]
[3]
[4]
[5]]
2. np.tile
函数原型：numpy.tile(A,reps) #简单理解是此函数将A进行重复输出
其中A和reps都是array_like的参数，A可以是：array,list,tuple,dict,matrix以及基本数据类型int,string,float以及bool类型,reps的类型可以是tuple,list,dict,array,int,bool,但不可以是float,string,matrix类型。
计较常用的形式有两种，是将A简单进行一维重复输出，和将A进行二维重复后输出。
一维重复：
import numpy as np
a = [[1,2,3],[4,5,5]]
b = np.tile(a,3)
print(b)
#输出为
#[[1 2 3 1 2 3 1 2 3]
# [4 5 5 4 5 5 4 5 5]]二维重复：#上面的一维重复相当于 b = np.tile(a,[1,3])
import numpy as np
a = [[1,2,3],[4,5,5]]
b = np.tile(a,[2,3])
print(b)
#输出为：
#[[1 2 3 1 2 3 1 2 3]
# [4 5 5 4 5 5 4 5 5]
# [1 2 3 1 2 3 1 2 3]
# [4 5 5 4 5 5 4 5 5]]2.1 np.tile
numpy.tile()是个什么函数呢，说白了，就是把数组沿各个方向复制
比如 a = np.array([0,1,2]), np.tile(a,(2,1))就是把a先沿x轴（就这样称呼吧）复制1倍，即没有复制，仍然是 [0,1,2]。 再把结果沿y方向复制2倍，即最终得到
array([[0,1,2],
[0,1,2]])
同理：
>>> b = np.array([[1, 2], [3, 4]]) >>> np.tile(b, 2) #沿X轴复制2倍 array([[1, 2, 1, 2], [3, 4, 3, 4]]) >>> np.tile(b, (2, 1))#沿X轴复制1倍（相当于没有复制），再沿Y轴复制2倍 array([[1, 2], [3, 4], [1, 2], [3, 4]])
numpy.tile()具体细节，如下：
Construct an array by repeating A the number of times given by reps.
If reps has length d, the result will have dimension of max(d, A.ndim).
If A.ndim < d, A is promoted to be d-dimensional by prepending new axes. So a shape (3,) array is promoted to (1, 3) for 2-D replication, or shape (1, 1, 3) for 3-D replication. If this is not the desired behavior, promote A to d-dimensions manually before calling this function.
If A.ndim > d, reps is promoted to A.ndim by pre-pending 1’s to it. Thus for an A of shape (2, 3, 4, 5), a repsof (2, 2) is treated as (1, 1, 2, 2).
Note : Although tile may be used for broadcasting, it is strongly recommended to use numpy’s broadcasting operations and functions.
| Parameters: | A : array_like
reps : array_like
|
|---|---|
| Returns: | c : ndarray
|
See also
- repeat
- Repeat elements of an array. broadcast_to
- Broadcast an array to a new shape
Examples
>>> a = np.array([0, 1, 2])
>>> np.tile(a, 2)
array([0, 1, 2, 0, 1, 2])
>>> np.tile(a, (2, 2))
array([[0, 1, 2, 0, 1, 2],
[0, 1, 2, 0, 1, 2]])
>>> np.tile(a, (2, 1, 2))
array([[[0, 1, 2, 0, 1, 2]],
[[0, 1, 2, 0, 1, 2]]])
>>> b = np.array([[1, 2], [3, 4]])
>>> np.tile(b, 2)
array([[1, 2, 1, 2],
[3, 4, 3, 4]])
>>> np.tile(b, (2, 1))
array([[1, 2],
[3, 4],
[1, 2],
[3, 4]])
>>> c = np.array([1,2,3,4])
>>> np.tile(c,(4,1))
array([[1, 2, 3, 4],
[1, 2, 3, 4],
[1, 2, 3, 4],
[1, 2, 3, 4]])