Learning What and Where to Draw

Introduction

Recently Generative Adversarial Networks (GANs) have been highly successful in several machine learning applications. Specifically, these models have been successfully used by to synthesize real-world images. GANS consist of two components: a generator network and a discriminator network. The objective of the generator is to synthesize images that the discriminator will classify as real. In turn, the objective of the discriminator is to classify it's input as synthetic or real. In what follows I outline the contents of the paper 'Learning What and Where to Draw' by Akata et al. (2016).