R Deep Learning Essentials - Second Edition

Generative adversarial networks

Generative Adversarial Networks have been called the coolest thing since sliced bread by Yann LeCunn, one of the most prominent people in deep learning. If he believes that, then we should all take notice!

Most of our models in this book have been discriminative models, that is, we try to differentiate one class from another. However in Chapter 9, Anomaly Detection and Recommendation Systems we created a generative model in the anomaly detection use-case. This model could create new data, albeit a different representation of the input data. Creating complex generative models is a very hot research topic in deep learning. Many believe that generative models can solve many problems in deep learning, including ...