generative

In deep learning, Variational autoencoders (VAEs) is a technique for learning latent representations. They are also used in a number of semi supervised tasks such as to draw/generate images.