A Framework for Unsupervised Deep Clustering

Not only can clustering in itself be a useful task to solve, features learnt from deep clustering algorithms can also be used in downstream tasks such as sleep stage classification. Using an open dataset such as the Temple University Hospital EEG corpus, I plan to develop a method for Deep Clustering of EEG data based on autoencoders. Having previously worked with this dataset, I am familiar with it's processing and using the processed data for Transformer pretraining. In addition to this, I am also familiar with general signal processing as I am an Electrical Engineering Masters student. At the end of the program I would hope to have developed an opensource method which can be published and which can be finetuned for other downstream tasks.