Currently, there are many solutions for spatiotemporal sequence prediction problems, such as ConvLSTM, SocialLSTM, etc.
However, in my personal opinion, these algorithms are very difficult to understand and not user-friendly for beginners or non-professionals in machine learning.
I have no background in machine learning and do not want to delve into those algorithms. I just want to use the scenario I want.
Personally, I feel that this problem is simplified and should not be difficult for an algorithm.
The simplified problem is as follows: There is a plane with 9 points on it, each representing a data point that is known. A set of nine data points represents the state at one time, and the data for n time steps (t-n…t-1, t) is known. I want to predict the state for future time steps.
Please design some simpler and more understandable algorithms to solve this problem. The simpler the algorithm, the better.
