
PhD
Efficient Learning and Probabilistic Inference for Science (ELPIS) at TUM and Helmholtz AI
Ingolstädter Landstraße 1
85764 Oberschleißheim
Biosketch
PhD candidate in Bayesian Machine Learning. Previously in Finance. Academic background Maths and Computer Science (St. John’s College, Oxford)
relAI Research
More efficient yet reliable learning through Bayes
Creating models with quantifiable and correct uncertainty estimates is a challenging open problem with connections to multiple distinct areas of machine learning.
The mathematical framework of Bayesian statistics offers a principled and provably optimal way of approaching this problem. However, naive Bayesian algorithms require a lot of resources and are in many cases intractable.
Specialized algorithms and models are therefore necessary for applying Bayes in practice. This is the broad field wherein my research lies.
More concretely, I am interested in alternative formulations of Bayes, continual and meta learning. A unifying thread behind these interests is learning under change. My research aims to design algorithms that do not silently break down when applied to different circumstances than those they were originally trained on.