MENU

Andreas Krause

Andreas Krause
Prof. Dr.
Andreas Krause
PI
ETH Zurich
krausea@ethz.ch
Curriculum Vitae
I’m excited about interdisciplinary research connecting machine learning and control theory and opportunities for impact in major application domains.

Andreas Krause is a Professor of Computer Science at ETH Zurich, where he leads the Learning & Adaptive Systems Group in the Institute for Machine Learning at ETH. He also serves as Academic Co-Director of the Swiss Data Science Center and Chair of the ETH AI Center. He received his Ph.D. in Computer Science from Carnegie Mellon University (2008) and his Diplom in Computer Science and Mathematics from the Technical University of Munich, Germany (2004).

Scientific Publications

Published
Learning Long-Term Crop Management Strategies with CyclesGym
36th Conference on Neural Information Processing Systems (NeurIPS 2022)
Matteo Turchetta, Luca Corinzia, Scott Sussex, Amanda Burton, Juan Herrera, Ioannis Athanasiadis, Joachim Buhmann, Andreas Krause
Collection
Published
Near-Optimal Multi-Agent Learning for Safe Coverage Control
36th Conference on Neural Information Processing Systems (NeurIPS 2022)
Manish Prajapat, Matteo Turchetta, Melanie Zeilinger, Andreas Krause
Collection
Published
Meta-Learning Priors for Safe Bayesian Optimization
Proceedings of The 6th Conference on Robot Learning
Pages 205-237
Jonas Rothfuss, Christopher Koenig, Alisa Rupenyan, Andreas Krause
Collection Arxiv.org
Published
Model-based Causal Bayesian Optimization
ArXiv
Scott Sussex, Anastasiia Makarova, Andreas Krause
Collection Arxiv.org
Automatic termination for hyperparameter optimization
arxiv
Anastasiia Makarova, Huibin Shen, Valerio Perrone, Aaron Klein, Andreas Krause, Matthias Seeger, Cedric Archambeau
Doi.org Collection Arxiv.org
Published
GoSafeOpt: Scalable Safe Exploration for Global Optimization of Dynamical Systems
Artificial Intelligence
Vol 320 Pages 103922
Bhavaya Sukhija, Matteo Turchetta, David Lindner, Andreas Krause, Sebastian Trimpe, Dominik Baumann
Collection
Published
Risk-averse Heteroscedastic Bayesian Optimization
Advances in Neural Information Processing Systems
Vol 34 Pages 17235--17245
Anastasiia Makarova, Ilnura Usmanova, Ilija Bogunovic, Andreas Krause
Doi.org Collection
Published
Robust Generalization despite Distribution Shift via Minimum Discriminating Information
Conference on Neural Information Processing Systems (NeurIPS 2021)
Vol 34 Pages 29754-29767
Tobias Sutter, Andreas Krause, Daniel Kuhn
Doi.org Collection
Published
Distributional Gradient Matching for Learning Uncertain Neural Dynamics Models
Conference on Neural Information Processing Systems (NeurIPS 2021)
Pages 29780 - 29793
Lenart Treven, Philippe Wenk, Florian Dörfler, Andreas Krause
Doi.org Collection
Published
Near-Optimal Multi-Perturbation Experimental Design for Causal Structure Learning
Advances in Neural Information Processing Systems
Vol 34 Pages 777--788
Scott Sussex, Andreas Krause, Caroline Uhler
Doi.org Collection
Learning Stable Deep Dynamics Models for Partially Observed or Delayed Dynamical Systems
Conference on Neural Information Processing Systems (NeurIPS 2021)
Andreas Schlaginhaufen, Philippe Wenk, Andreas Krause, Florian Dörfler
Collection
Safe and Efficient Model-free Adaptive Control via Bayesian Optimization
IEEE International Conference on Robotics and Automation (ICRA 2021)
Christopher Koenig, Matteo Turchetta, John Lygeros, Alisa Rupenyan, Andreas Krause
Collection
PopSkipJump: A Decision-based Adversarial Attack for Probabilistic Classifiers
38th International Conference on Machine Learning (ICML 2021)
Vol 139 Pages 9712 – 9721
Carl-Johann Simon-Gabriel, Noman Ahmed Sheikh, Andreas Krause
Collection

Research projects

Title
Principal Investigators

Safe model-based reinforcement learning via causal inference and meta-learning

Andreas Krause

Summary

A key limitation of current deep reinforcement learning (RL) approaches is their need for accurate computational models and focus on fully observable domains.  In many real-world domains, such as those considered in the NCCR, only approximate dynamics models exist, and partial observability is paramount. In our research, we will investigate connections between reinforcement learning and causal inference. We explore novel approaches for off-policy evaluation based on ideas from causal inference, and study the use of experimental design for causal discovery for safe exploration in RL. We plan to also utilize ideas from meta-learning in order to transfer inductive biases across multiple related tasks.

People involved

Andreas Krause
Simon-Gabriel Carl-Johann
Andreas Krause, Matteo Turchetta, , Carl-Johann Simon-Gabriel
Safe model-based reinforcement learning via causal inference and meta-learning
Andreas Krause

A key limitation of current deep reinforcement learning (RL) approaches is their need for accurate computational models and focus on fully observable domains.  In many real-world domains, such as those considered in the NCCR, only approximate dynamics models exist, and partial observability is paramount. In our research, we will investigate connections between reinforcement learning and causal inference. We explore novel approaches for off-policy evaluation based on ideas from causal inference, and study the use of experimental design for causal discovery for safe exploration in RL. We plan to also utilize ideas from meta-learning in order to transfer inductive biases across multiple related tasks.

Andreas Krause
Simon-Gabriel Carl-Johann
Andreas Krause, Matteo Turchetta, , Carl-Johann Simon-Gabriel
144
3c40964f-146e-4bac-bb7e-eefb82eb4a46

Distributed GP learning and model-based RL

Andreas Krause, Giancarlo Ferrari Trecate

Summary

The goal of the project is to study data-driven modelling and control frameworks for large-scale systems stemming from the interconnection of several subsystems, which are of primary interest to the NCCR. For this purpose, we will develop (i) innovative probabilistic machine learning algorithms, based on Gaussian Processes (GPs), for batch and online estimation of models complying with the coupling structure of the system and (ii) model-based Distributed Reinforcement Learning (DRL) algorithms for the generation of networked control policies. Our DRL schemes will exploit the epistemic model uncertainty to drive exploration while ensuring stability of the distributed controllers.

People involved

Andreas Krause
Giancarlo Ferrari Trecate
Mahrokh Ghoddousi Boroujeni
Andreas Krause, Giancarlo Ferrari Trecate, Mahrokh Ghoddousi Boroujeni
Distributed GP learning and model-based RL
Andreas Krause, Giancarlo Ferrari Trecate

The goal of the project is to study data-driven modelling and control frameworks for large-scale systems stemming from the interconnection of several subsystems, which are of primary interest to the NCCR. For this purpose, we will develop (i) innovative probabilistic machine learning algorithms, based on Gaussian Processes (GPs), for batch and online estimation of models complying with the coupling structure of the system and (ii) model-based Distributed Reinforcement Learning (DRL) algorithms for the generation of networked control policies. Our DRL schemes will exploit the epistemic model uncertainty to drive exploration while ensuring stability of the distributed controllers.

Andreas Krause
Giancarlo Ferrari Trecate
Mahrokh Ghoddousi Boroujeni
Andreas Krause, Giancarlo Ferrari Trecate, Mahrokh Ghoddousi Boroujeni
110
63103095-1b2c-42bc-96bf-4e514b510220

Distributionally robust reinforcement learning

Daniel Kuhn, Andreas Krause

Summary

This project uses ideas from mathematical optimization and learning theory to derive statistical guarantees for reinforcement learning (RL). While there has recently been significant progress in the understanding of finite-sample guarantees for the linear quadratic regulator (LQR) problem, it remains unclear how to generalize these results beyond the LQR setting. We plan to establish a principled approach to RL based on distributionally robust optimization with the goal to derive statistical guarantees for RL problems beyond the LQR setup. We also aim to develop efficient computational algorithms.

People involved

Daniel Kuhn
Andreas Krause
Tobias Sutter
Daniel Kuhn, Andreas Krause, Tobias Sutter
Distributionally robust reinforcement learning
Daniel Kuhn, Andreas Krause

This project uses ideas from mathematical optimization and learning theory to derive statistical guarantees for reinforcement learning (RL). While there has recently been significant progress in the understanding of finite-sample guarantees for the linear quadratic regulator (LQR) problem, it remains unclear how to generalize these results beyond the LQR setting. We plan to establish a principled approach to RL based on distributionally robust optimization with the goal to derive statistical guarantees for RL problems beyond the LQR setup. We also aim to develop efficient computational algorithms.

Daniel Kuhn
Andreas Krause
Tobias Sutter
Daniel Kuhn, Andreas Krause, Tobias Sutter
106
7e6307b6-fce1-4184-b451-bb44a15e437a

Continuous-time Deep Model-based Reinforcement Learning

Florian Dörfler, Andreas Krause

Summary

We investigate how classical methods from system identification, dynamical systems, and control can be combined with techniques from non- parametric machine learning to develop a rigorous framework for continuous-time model-based deep reinforcement learning. Concretely, we will learn a continuous-time dynamics model via neural ordinary differential equations, along with associated uncertainty bounds that can be used to trade exploration and exploitation. Ultimately, our framework should enable control design for complex and uncertain systems, and we plan to apply it to problems in building automation and grid-connected converters.

People involved

Florian Dörfler
Andreas Krause
Lenart Treven
Florian Dörfler, Andreas Krause, Lenart Treven
Continuous-time Deep Model-based Reinforcement Learning
Florian Dörfler, Andreas Krause

We investigate how classical methods from system identification, dynamical systems, and control can be combined with techniques from non- parametric machine learning to develop a rigorous framework for continuous-time model-based deep reinforcement learning. Concretely, we will learn a continuous-time dynamics model via neural ordinary differential equations, along with associated uncertainty bounds that can be used to trade exploration and exploitation. Ultimately, our framework should enable control design for complex and uncertain systems, and we plan to apply it to problems in building automation and grid-connected converters.

Florian Dörfler
Andreas Krause
Lenart Treven
Florian Dörfler, Andreas Krause, Lenart Treven
99
2d302258-eb79-48e4-a277-d88660d68bfa

Risk aware Bayesian optimization with applications to adaptive control

Alisa Rupenyan, Andreas Krause

Summary

Our goal is to develop novel algorithms for risk-aware Bayesian optimization and demonstrate them in a case study for high-precision linear drives requiring position accuracy and stability with less than 5 nm deviation. For this system, as for many other applications, it is crucial to optimize expected performance and ensure reliable outcomes.  We will develop computationally efficient algorithms trading expected performance with the variance in the outcome, considering risk measures on the optimization constraints, and compiling thi acquisition functions into neural network policies.

People involved

Alisa Rupenyan
Andreas Krause
Alisa Rupenyan, Andreas Krause
Risk aware Bayesian optimization with applications to adaptive control
Alisa Rupenyan, Andreas Krause

Our goal is to develop novel algorithms for risk-aware Bayesian optimization and demonstrate them in a case study for high-precision linear drives requiring position accuracy and stability with less than 5 nm deviation. For this system, as for many other applications, it is crucial to optimize expected performance and ensure reliable outcomes.  We will develop computationally efficient algorithms trading expected performance with the variance in the outcome, considering risk measures on the optimization constraints, and compiling thi acquisition functions into neural network policies.

Alisa Rupenyan
Andreas Krause
Alisa Rupenyan, Andreas Krause
98
09023dea-22c0-46b6-aa47-70bfde44c173