Jakob Heiss

Since 2024, I have been doing my postdoc in the Yu group at UC Berkeley with Prof. Bin Yu on Deep Learning Theory and Uncertainty Quantification.

2019-2024, I did my Ph.D. advised by Prof. Josef Teichmann in the Stochastic Finance Group at ETH Zurich, and was affiliated with the ETH AI Center.

My main research interest is the mathematical theory of deep learning algorithms (in terms of their inductive bias, multi-task learning, and compressability). Additionally, I work on quantifying the epistemic and aleatoric uncertainty of deep neural networks, and I apply deep learning to market design (preference elicitation for combinatorial auctions). I also work on Neural Jump ODEs for irregularly observed time series and on compression of neural networks.

Previously, I received a B.Sc. and M.Sc. (2019) in Technical Mathematics from the Technical University of Vienna.

What do I enjoy about being a researcher?

I love working in teams on scientifically exciting open questions, trying to understand paradoxical phenomena. I particularly like to establish new perspectives on such questions, which can partially resolve such paradoxes. I also enjoy developing new methods and diagnosing, discussing, understanding, and improving methods in teams. It is important to me that each team member is passionate about the projects, and I am happy when students continue to work with me in their free time for years after my supervision purely out of excitement about our work.

News

December 17, 2025, at INCHER & Zell Colloquium, I give an online lecture on "The inductive bias of deep neural networks toward feature learning".
October 13, 2025, at the World Leading Scientists 2025 Conference 2025 at Berkeley, I present a poster on Uncertainty Quantification.
October 3-4, 2025, at the Berkeley–Columbia Meeting in Engineering and Statistics at Berkeley, I give a talk on "Machine Learning for Preference Elicitation in Combinatorial Auction Design" (October 4)
September 25-26, 2025, at MoDL in New York, I present a poster on Deep Learning Theory
July 17, 2025, at ICML 2025 in Vancouver, I give an oral presentation (top 1%) of our paper "Prices, Bids, Values: One ML-Powered Combinatorial Auction to Rule Them All". [12' presentation] [5' video]
February 16-21, 2025, I participate in the workshop on Uncertainty Quantification in Neural Network Models at BIRS in Banff (Canada), giving a talk and leading a discussion on "Inductive Bias of Neural Networks" and presenting a poster on Uncertainty Quantification.
October 1, 2024, I start my postdoc with Prof. Bin Yu in the Yu group at UC Berkeley on Deep Learning Theory and Uncertainty Quantification.
July 26, 2024, I defend my PhD thesis on "Inductive bias of neural networks and selected applications".
July 8-12, 2024, I am at the 12th Bachelier World Congress of the Bachelier Finance Society in Rio de Janeiro, giving a talk on "Path-dependent Neural Jump ODEs and their Application to Stochastic Filtering".
June 17-20, 2024, I am at the ETH – Hong Kong – Imperial Mathematical Finance Workshop at Imperial College London, giving a talk on "Deep Learning Theory on Multi-task Learning".
On March 25, 2024, I am at AMLD EPFL 2024 in Lausanne to give a talk on "NOMU: Neural Optimization-based Model Uncertainty"
From Feb 19-26, 2024, I am at AAAI 2024 in Vancouver for our paper "Machine Learning-powered Combinatorial Clock Auction".
From Jun 25-28, 2023, I am at the University of Oxford.
From Feb 15-28, 2023, I am visiting Bin Yu's research group at UC Berkeley.
From Feb 6-15, 2023, I am at AAAI 2023 in Washington, DC, to give an oral presentation of our paper "Bayesian Optimization-based Combinatorial Assignment". A pre-recorded video of the talk is on YouTube: https://youtu.be/6YH9K6LDHPY.

Page updated

Google Sites

Report abuse