解析mxnet.ndarray.reshape的用法，和numpy.reshape很像

1，常规用法。

>>> import mxnet.ndarray as nd
>>> a = nd.reshape(nd.arange(12), shape=(3, 4))  # shape=(3,4)可直接写（3,4），不过我喜欢把实参名标上。
>>> a
[[  0.   1.   2.   3.]
 [  4.   5.   6.   7.]
 [  8.   9.  10.  11.]]
<NDArray 3x4 @cpu(0)>

2，利用特殊值 {0, -1, -2, -3}来表现形状。

　　(1),0表示从输入复制维度，如：

- input shape = (2,3,4), shape = (4,0,2), output shape = (4,3,2)  # 有0的位置，维度不变
- input shape = (2,3,4), shape = (2,0,0), output shape = (2,3,4)  # 有0的位置，维度不变

　　(2),-1所在位置用来保持输入和输出矩阵的总数量不变。如：

>>> import mxnet.ndarray as nd
>>> a = nd.reshape(nd.arange(12), shape=(3, -1))  # -1这个位置会被自动计算为4，因为要保持和前面的总数量一样
[[  0.   1.   2.   3.]
 [  4.   5.   6.   7.]
 [  8.   9.  10.  11.]]
<NDArray 3x4 @cpu(0)>

　　(3),-2表示复制所有或者剩下的维度。如：

- input shape = (2,3,4), shape = (-2,), output shape = (2,3,4)  # 复制所有维度
- input shape = (2,3,4), shape = (2,-2), output shape = (2,3,4)  # 复制第一维以为的维度
- input shape = (2,3,4), shape = (-2,1,1), output shape = (2,3,4,1,1)  # 复制最后两维以外的维度

　　(4),-3表示将连续两个维度乘积作为新的维度。如：

- input shape = (2,3,4), shape = (-3,4), output shape = (6,4)
- input shape = (2,3,4,5), shape = (-3,-3), output shape = (6,20)
- input shape = (2,3,4), shape = (0,-3), output shape = (2,12)
- input shape = (2,3,4), shape = (-3,-2), output shape = (6,4)