A generative adversarial network (GAN) is a class of machine learning frameworks conceived in 2014 by Ian Goodfellow and his colleagues. Two neural networks (Generator and Discriminator) compete with each other like in a game. This technique learns to generate new data using the same statistics as that of the training set, given a training set.

Generated Images

In this post, we’ll train a GAN to generate images of cats’ faces. We will use the Cats Faces Dataset which consists of more than 15,700 cats images. Since generative modeling is an unsupervised learning method, hence there are no labels on the images.

#data-science #pytorch #artificial-intelligence #machine-learning #gans