np.newaxis

np.newaxis----增加矩阵维度

常用方法--将（10，）维度变为固定维度（10，1）的矩阵

x=np.linspace(-0.5,0.5,10)
print(x)
print(x.shape)

[-0.5        -0.38888889 -0.27777778 -0.16666667 -0.05555556  0.05555556
  0.16666667  0.27777778  0.38888889  0.5       ]
(10,)

p=x[:,np.newaxis]
print(p)
print(p.shape)

[[-0.5       ]
 [-0.38888889]
 [-0.27777778]
 [-0.16666667]
 [-0.05555556]
 [ 0.05555556]
 [ 0.16666667]
 [ 0.27777778]
 [ 0.38888889]
 [ 0.5       ]]
(10, 1)

高维度

a=np.random.rand(4,4)
print(a)

[[ 0.61690279  0.13492536  0.77801097  0.27768464]
 [ 0.17720794  0.55726312  0.86292334  0.24805834]
 [ 0.29273676  0.97337849  0.0301729   0.89382466]
 [ 0.88650348  0.6262905   0.42488666  0.80979559]]

增加所有维度

d=a[:,np.newaxis]
print(d)
print(d.shape)

[[[ 0.61690279  0.13492536  0.77801097  0.27768464]]

 [[ 0.17720794  0.55726312  0.86292334  0.24805834]]

 [[ 0.29273676  0.97337849  0.0301729   0.89382466]]

 [[ 0.88650348  0.6262905   0.42488666  0.80979559]]]
(4, 1, 4)

增加部分维度

b=a[:2,np.newaxis]
print(b)
print(b.shape)

[[[ 0.61690279  0.13492536  0.77801097  0.27768464]]

 [[ 0.17720794  0.55726312  0.86292334  0.24805834]]]
(2, 1, 4)

c=a[:,:2,np.newaxis]
print(c)
print(c.shape)

[[[ 0.61690279]
  [ 0.13492536]]

 [[ 0.17720794]
  [ 0.55726312]]

 [[ 0.29273676]
  [ 0.97337849]]

 [[ 0.88650348]
  [ 0.6262905 ]]]
(4, 2, 1)

 

 

​