As a new learner of Numpy, it is very common to be confused by the form of array, braces nested in braces, such as ‘a= np.array[[[1],[2],[3]]]’ so that its shape cannot be precisely known by the users, although the shape can be fetched through ‘a.shape’, making users unable to perform multidimensional array correctly.
The method is as the following：(1)every array needs a most out brace.(2) Go inner, the number of paired brace is the element number of 0 axis, 1 axis, 2 axis etc.(3) The dimension number increased along with the going-inner, until the number element is reached, the sum of levels is the dimensional number.
For example:
>>> import numpy as np >>> data=np.arange(4) >>> data1=data.reshape((4,1)) >>> data2=data.reshape((1,4)) >>> data3=data.reshape((2,2)) >>> data4=data.reshape((1,2,2)) >>> data5=data.reshape((2,1,2)) >>> data6=data.reshape((2,2,1)) >>> data array([0, 1, 2, 3]) >>> data1 array([[0], [1], [2], [3]]) >>> data2 array([[0, 1, 2, 3]]) >>> data3 array([[0, 1], [2, 3]]) >>> data4 array([[[0, 1], [2, 3]]]) >>> data5 array([[[0, 1]], [[2, 3]]]) >>> data6 array([[[0], [1]], [[2], [3]]])
①data1's shape is (4,1),so firstly, the most out brace pair shall be set-->[ ], and then 0 axis is 4, meaning 4 pairs of brace inside -->[ [ ],[ ],[ ],[ ] ] ,finally, 1 axis is 1,and is the final dimension, meaning 1 number element shall be inside-->[ [0],[1],[2],[3]].
② data2's shape is (1,4), firstly-->[ ], then -->[ [ ] ], finally -->[ [ 0,1,2,3] ]
③ data3's shape is (2,2), firstly-->[ ] ,then--> [ [ ], [ ] ], finally-->[ [0,1], [2,3] ]
④data4' shape is (1,2,2), firstly-->[ ] ,then-->[ [ ] ], then--> [ [ [ ], [ ] ] ], finally--> [ [ [ 0,1], [2,3 ] ] ]
⑤data5's shape is (2,1,2),firstly--> [ ],then-->[ [ ],[ ] ], then-->[ [ [ ] ], [ [ ] ] ], finally --> [ [ [ 0,1],[ [ 2,3] ] ]
⑥data6's shape is (2,2,1),firstly-->[ ],then --> [ [ ] ,[ ] ],then-->[ [ [ ], [ ] ], [ [ ] , [ ] ] ],finally-->[ [ [ 0],[1] ], [ [2],[3] ] ]