First test. Gan learning using noise input, no training data. Patterns are lovely (+ useful too) but probably I have to use tiny amount of "real" image data.
Anyway, I will continue experimenting today with more richer "noise" input, using waveforms etc

#mlsky #aisky