Neural network for generating matrices?

Question

Python

Good afternoon. I would like to write a neural network that, based on a dataset (1500+ matrices), should generate similar ones.
Those. the neural network must understand what is the relationship between the numbers in the matrix, and generate a similar one. Tell me where to dig?

Answer 1

Well let's think.
If you have 1500 matrices of the same size, like 9x12 - that's fine, no need to puzzle over the choice of inputs, here are 108 ready-made.
But what if all the matrices are of different sizes? How to remove the input signal from them? Image networks handle this easily by scaling images to the same size. But your neural network

must understand what is the relationship between the numbers in the matrix,

those. numbers, apparently, it is undesirable to distort the same algorithms as in the processing of pictures.
I don’t see a good solution, but I’m too lazy to paint a bad one. I already scribbled four times more than your question.
Dig in the direction of "think".