This is a talk given by Ole Winther as part of the weekly Science Seminar Series at DTU Biosustain (Techncial University of Denmark).
Pre-training deep learning models on large amounts of unlabelled data, for example from the internet or UniProt, has in the last few years turned out to be a very powerful foundation for building powerful AI models. In this talk, Ole will explain the basic concepts, the perspectives in the medium horizon and give some examples from his groups works (https://arxiv.org/abs/2207.08143, https://www.biorxiv.org/content/10.1101/2022.04.08.487609v1 and more).
DTU Biosustain, Technical University of Denmark
DTU Microbes Initiative, Technical University of Denmark
Contact: Sünje Johanna Pamp (email@example.com)