高效重塑3Dnumpy数组

Question

假设我有一个2Dnumpy数组，并想要将其步进到3D，那么最好的方法是什么？

小例子：

def find_ngrams(input_list, n):
    return np.array(list(zip(*[input_list[i:] for i in range(n)])))

x = np.array(range(15))
x = x.reshape((5,3))
print(x)
print(x.shape)

res = find_ngrams(x, 3)
print(res.shape)
print(res)

这将正确返回预期结果：

[[ 0  1  2]
 [ 3  4  5]
 [ 6  7  8]
 [ 9 10 11]
 [12 13 14]]
(5, 3)
(3, 3, 3)
[[[ 0  1  2]
  [ 3  4  5]
  [ 6  7  8]]

 [[ 3  4  5]
  [ 6  7  8]
  [ 9 10 11]]

 [[ 6  7  8]
  [ 9 10 11]
  [12 13 14]]]

但是，我如何才能更有效地执行此操作，最好是使用stride_tricks

Answer 1

下面是我如何使用as_strided实现这一点

window_length=3

strides = x.strides

new_len = (x.shape[0]-window_length+1)

out = as_strided(x,shape=(window_length, new_len, x.shape[1]),
                 strides=(strides[0],) + (strides[0], strides[1]))

输出：

array([[[ 0,  1,  2],
        [ 3,  4,  5],
        [ 6,  7,  8]],

       [[ 3,  4,  5],
        [ 6,  7,  8],
        [ 9, 10, 11]],

       [[ 6,  7,  8],
        [ 9, 10, 11],
        [12, 13, 14]]])