This unit focuses on advanced deep learning concepts and their application. Assuming basic machine and deep learning knowledge, and experience in related programming frameworks, it will delve deeper into the building blocks of modern deep learning systems and their specialised applications in processing structured and unstructured data. The unit covers in detail advanced topics in Generative Adversarial Networks (GANs), variational autoencoders, deep reinforcement learning, policy gradient methods, and Adversarial Machine Learning. These topics are covered with hands-on experience throughout the unit.

Students are able to (1) apply deep neural networks to structured and unstructured data; (2) apply generative adversarial networks to learn data distribution; (3) analyse generative models for cross-domain data generation; (4) demonstrate understanding of reinforcement learning; (5) compute adversarial patterns for deep models; and (6) contrast robust deep models with non-robust models.