Objectives and outcomes
Introducing deep learning concepts, theoretical overview and the implementation of various deep learning algorithms through projects. After this course, students will be able to recognise problems that can be solved with deep learning. They will be able to read most of the state-of-the-art research papers on deep learning and understand the mathematical and technical solutions presented.
Introduction, a modular approach to learning deep architectures. Practical aspects (normalisation, regularisation, hyperparameter tuning). Optimisation algorithms. Deep neural networks. Deep convolutional neural networks – different architectures, practical aspects, object detection, face recognition and style transfer). Recurrent neural networks – LSTM (long-short term memory), applications such as speech recognition and natural language processing, sequence-to-sequence models. Reinforcement learning. Generative Adversarial Networks.
Implementation of deep learning algorithms covered in lectures. Introduction. Neural networks. Model selection and evaluation. Practical aspects of deep learning. Optimisation algorithms. Deep convolutional neural networks and recurrent neural networks. Deep generative models – generative adversarial networks, autoencoders. Reinforcement learning.