Sustainable representation learning with causal ML
Representation learning approaches provide remarkable results but are often based on complex architectures requiring huge datasets and unsustainable computations. Despite this gargantuan amount of data and compute, state-of-the-art methods often generalize poorly in practice due to slight changes in between the training and test distribution. Causal machine learning offers an elegant approach to study this problem but it does not scale as well as modern deep learning. Our goal in this project is to explore how causal structure in the data distribution can inform the design of novel forms of inductive bias, making AI models easier to train, more sustainable, and more reliable in solving real world problems.
The project will be co-supervised by Nicoletta Noceti and Lorenzo Rosasco (UniGe) and Francesco Locatello (Amazon Web Services). The student will be part of the prestigious ELLIS PhD program (European Laboratory for Learning and Intelligent System -- https://ellis.eu/), allowing a unique exposure to leading research environments in both academia and industry. The student will gain valuable experience in both, sharing their time between the Genova machine learning center (MaLGa) and the Amazon AWS research labs in Tuebingen, Germany. Competitive compensation applies to support the stays abroad.
Towards Causal Representation Learning https://arxiv.org/abs/2102.11107
Desiderata For Representation Learning: A Causal Perspective https://arxiv.org/abs/2109.03795
Representation Learning via Invariant Causal Mechanisms https://arxiv.org/abs/2010.07922
Topics: Computer Vision, Machine Learning.
Perspective candidates should make an expression of interest before 22/9/2021.
Full application will be available at http://phd.dibris.unige.it/csse/index.php/admission-procedure