Research description
Machine learning (ML) is the key technology of our time. However, so far, its success crucially relies on human machine learning experts to perform manual tasks, thus limiting its potential impact. To overcome this issue, here at the ML Lab, we focus on the progressive automation of machine learning (AutoML) in order to democratize access to ML by making state-of-the-art ML solutions accessible for everyone, in world-leading open source systems. Framed differently, and from a technical point of view, our lab develops AI that builds and improves AI.
Particular technical fields of interest for us include meta-learning, neural architecture search, efficient hyperparameter optimization, deep learning for tabular data, multi-objective optimization, proper benchmarking, and AutoRL. We also apply AutoML to improve practical deep learning for several applications of high societal importance, such as RNA folding and design, EEG decoding, and data from the medical sector in general.