flatten multiple slidingwindow

[rapus95]

trafficstars

My current problem is that I want to predict some transformation result of an ill-posed problem (deconvolution). Thus I have a lot of input function plots (x,y-arrays) and their convolutions (with a given kernel) as prediction targets. Right now I implemented this training run as sliding window for the input data and a single value point from the output data which shall have the same index as the center of the sliding window.

The question I now have, is whether there is any built-in solution to "append" each of those (window,targetpoint) pairs of different function graphs in such a way together that I can shuffle them across function graphs?