Develops a machine learning model that combines a probabilistic variational autoencoder (VAE) with adversarial components from generative adversarial networks (GAN) and supervised learning that can be used to augment human judgment for autos and other products where aesthetic design plays a major role in market acceptance.