In the recent years, with the increasing availability of computational power and possibility to leverage large amounts of data, Machine Learning and Deep Learning techniques have gained more and more popularity leading, especially when used in a supervised setting, to performance equaling or even exceding human capabilities for a variety of tasks (e.g Face Recogntion). The problem with supervised learning is that it usually requires humans to manually annotate the data and this process, particularly when performed on large datasets, is very time consuming and potentially expensive.

For this reason the focus of my reasearch is to explore novel Unsupervised Learning models and techniques which are able to automatically infer useful information and extract patterns from a given dataset without the need of any human supervision. This characteristic could potentially allow to leverage the large amount of uncurated data freely available on the internet in order to train deep learning models in a very time and cost efficient way.