what are the insights you were able to draw in terms of explanability?

[bhomass]

I went thru your entire code and reproduced the plots. Sure, you got a whole bunch of clusters of various sorts. But to your professed goal of explaining the prediction results from the neural network embedding, what insights exactly were you able to extract? It really escapes me what you see in any of the cluster diagrams.