为ltsm构造numpy数据（示例）

Question

我在理解如何为不同模型准备数据时遇到问题：

1. 一对多
2. 多对一
3. 多对多（A）
4. 多对多（B）



这是以这种方式思考的正确方法。形状编号无关紧要，与图片上的形状编号不匹配。我只是想了解背后的逻辑。：

import numpy as np


#1. one to many
# X for input y for output

X = np.ones([10,1,5])
y = np.zeros([10,3]) #3 represnts size of output vector

#2. many to one

X = np.ones([10,5,5])
y = np.zeros([10,1])

#3. many to many

X = np.ones([10,5,5])
y = np.zeros([10,5])  

# in this case cell should be different than y. It must be bigger to shift some data

#4. many to many

X = np.ones([10,5,5])
y = np.zeros([10,5])

# in this case cell is the same shape as y