[raverbashing]

Not everything requires a multiple layer NN to be trained. It is not optimal, sure, but it might open up different possibilities