Richard Feynman said “What I cannot create, I do not understand”. Given a domain of objects, say images, generative models try to ‘understand’ the domain well enough to be able to ‘create’ different objects from the domain, on their own or based on instructions. Generative adversarial networks (GANs) are the current hot favorites for the generative task, in part because Yann LeCun admitted that 'Adversarial training is the coolest thing since sliced bread.’ In this talk, we will try to overview the theory, practice and the hype around GANs and dissect the games GANs play.
Nishant Sinha is a Computer Science Researcher, Consultant and Mentor at [Kena Labs](http://www.kenalabs.com/). He is interested in building AI systems and solutions spanning textual, vision and speech modes, based on deep learning and logical reasoning. Previously, he held research positions at IBM Research and NEC Labs. He obtained his Ph.D. from Carnegie Mellon and B. Tech. from IIT Kharagpur.
Below is the video lecture:Written on July 24th, 2017 by