I am a Senior Data Scientist at the Swiss Data Science Center. In my research, I develop principled and practical algorithms for sequential decision-making. I also work on combining state-of-the-art deep learning and sequential decision-making methods in challenging real-world applications.
Before I completed a PostDoc with Csaba Szepesvári at the University of Alberta (supported by an Early Postdoc.Mobility fellowship of the Swiss National Science Foundation). I hold a PhD in machine learning that I completed under the supervision of Andreas Krause at ETH Zurich. In 2020, I completed a research internship at DeepMind. Prior, I received Bachelor and Master of Science degrees in Mathematics at ETH Zurich.
My research interests include reinforcement learning, adaptive experimental design, Bayesian optimization, robustness and safety, as well as deep learning and computer vision. I am equally excited about creating mathematical foundations for learning algorithms and bringing their value to real-world application. See below for some of my research highlights.
|Nov 1, 2023||
Two papers accepted at NeurIPS!
|Aug 21, 2023||I am starting a new position as a Sr. Data Sciencist at the Swiss Data Science Center in Zurich. Ping me if you like to catch up!|
|Aug 13, 2023||Our paper on Linear Partial Monitoring for Sequential Decision-Making: Algorithms, Regret Bounds and Applications, together with Tor Lattimore and Andreas Krause, got accepted for publication at JMLR.|
|May 1, 2023||I am helping to resurrect the reinforcement learning online seminars. Join us for exciting talks and engaging discussions!|
|Jan 20, 2023||
Two papers accepted:
|Feb 1, 2022||I am serving as Associate Chair at ICML 2022|
|Aug 1, 2021||Started my PostDoc at the University of Alberta|
|May 17, 2021||Successfully defended my PhD thesis!|
|Dec 15, 2020||I received an SNF Early PostDoc.Mobility Fellowship|
Safe Bayesian Optimization for Particle Accelerators
Together with collaboraters at PSI and ETH Zurich, I developed safe data-driven tuning algorithms for particle accelerators. Manually adjusting machine parameters is a re-occuring and time consuming task that is required on many acceletors and cuts down valuable time for experiments. A main difficulty is that all adjustments need to respect safety parameters to avoid damaging the machines (or trigger automated shutdown procedures). We successfully deployed our methods on two major experimental facilities at PSI, the High Intensity Proton Accelerator (HIPA) and the Swiss Free Electron Laser (SwissFEL).
Frequentist Analysis of Information-Directed Sampling
In my PhD thesis, I pioneered mathematical foundations of information-directed sampling (IDS), an algorithm design principle proposed by Daniel Russo and Benjamin Van Roy. Together with Tor Lattimore and Andreas Krause, I showed that the algorithm applies much more broadly to linear partial monitoring (and is provably near-optimal in all finite-action settings). More recently, I showed that IDS is also asymptotically optimal (together with Claire Vernade, Tor Lattimore and Csaba Szepesvári). This resolves an open problem in the literatue. It is also a remarkable result, because IDS was never explicitly designed for this regime.
Regret Minimization via Saddle Point OptimizationIn Proc. Neural Information Processing Systems (NeurIPS) Dec 2023
Linear Partial Monitoring for Sequential Decision-Making: Algorithms, Regret Bounds and ApplicationsJournal of Machine Learning Research (JMLR) Dec 2023
Managing Temporal Resolution in Continuous Value Estimation: A Fundamental Trade-offIn Proc. Neural Information Processing Systems (NeurIPS) Dec 2023
Near-optimal Policy Identification in Active Reinforcement LearningAccepted at ICLR (notable-top-5%) May 2023
Efficient Planning in Combinatorial Action Spaces with Applications to Cooperative Multi-Agent Reinforcement LearningAccepted at AISTATS Apr 2023
Tuning particle accelerators with safety constraints using Bayesian optimizationPhys. Rev. Accel. Beams Jun 2022
Information-Directed Sampling — Frequentist Analysis and ApplicationsJun 2021
Efficient Pure Exploration for Combinatorial Bandits with Semi-Bandit FeedbackIn Algorithmic Learning Theory Jun 2021
Bias-Robust Bayesian Optimization via Dueling BanditsIn Proc. International Conference on Artificial Intelligence and Statistics (AISTATS) Jul 2021
Asymptotically Optimal Information-Directed SamplingIn Proc. International Conference on Learning Theory (COLT) Aug 2021
Distributionally Robust Bayesian OptimizationIn Proc. International Conference on Artificial Intelligence and Statistics (AISTATS) Aug 2020
Information Directed Sampling for Linear Partial MonitoringIn Proc. International Conference on Learning Theory (COLT) Jul 2020
Experimental Design for Optimization of Orthogonal Projection Pursuit ModelsIn Proceedings of the 34th AAAI Conference on Artificial Intelligence (AAAI) Feb 2020
Bayesian Optimization for Fast and Safe Parameter Tuning of SwissFELIn Proc. International Free-Electron Laser Conference (FEL2019) Jun 2019
Adaptive and Safe Bayesian Optimization in High Dimensions via One-Dimensional SubspacesIn Proc. International Conference for Machine Learning (ICML) Jun 2019
Information-Directed Exploration for Deep Reinforcement LearningIn Proc. International Conference on Learning Representations (ICLR) May 2019
Stochastic Bandits with Context DistributionsIn Proc. Neural Information Processing Systems (NeurIPS) Dec 2019
Information Directed Sampling and Bandits with Heteroscedastic NoiseIn Proc. International Conference on Learning Theory (COLT) Jul 2018