Generative Adversarial Networks (GANS)

This week, we had two presentations: GANs (presented by Alan Zheng) and Style Transfer (presented by Jack Morris). We thought that the GANs paper was as good of a place as any for us to begin in terms of academic papers. It doesn’t get more fundamental than this. The original GANs paper outlines the novel setup of a GAN as a minimax game between a generator and discriminator. It contains some theory, which is complicated, but approachable (even for us!), as well as some really interesting applications– showing the amazing results you get when you sample from the latent space of a gan (generating new MNIST digits and pictures of faces).

Finally, just for kicks, we took a look at [] and had a group discussion about generated images, DeepFakes, and the problems these types of systems pose in today’s society.

Error! Looks like your browser can't display PDFs. Try using an updated version of Chrome or Firefox.