[{"id":15,"title":"Semi-Supervised Deep Learning","url":"https://staging.dresa.org.au/materials/semi-supervised-deep-learning.json","description":"Modern deep neural networks require large amounts of labelled data to train. Obtaining the required labelled data is often an expensive and time consuming process. Semi-supervised deep learning involves the use of various creative techniques to train deep neural networks on partially labelled data. If successful, it allows better training of a model despite the limited amount of labelled data available.\r\n\r\nThis workshop is designed to be instructor led and covers various semi-supervised learning techniques available in the literature. The workshop consists of a lecture introducing at a high level a selection of techniques that are suitable for semi-supervised deep learning. We discuss how these techniques can be implemented and the underlying assumptions they require. The lecture is followed by a hands-on session where attendees implement a semi-supervised learning technique to train a neural network. We observe and discuss the changing performance and behaviour of the network as varying degrees of labelled and unlabelled data is provided to the network during training.","doi":"","remote_updated_date":null,"remote_created_date":null,"scientific_topics":[],"operations":[]}]