generative-adversarial-model-experiments