Introduction

Deep Learning has brought significant progress in a variety of applications of machine learning in recent years. As powerful non-linear function approximators, their potential for use in learning-based control applications is very appealing. They benefit from large amounts of data, and present a very scalable solution e.g. for learning hard-to-model plant dynamics from data. Currently, the most widely- used method of training these deep networks are maximum likelihood approaches, which only give a point estimate of the parameters that maximize the likelihood of the input data, and do not quantify how certain the model is about its predictions. The uncertainty of the model is, however, a crucial factor in robust and risk-averse control applications. This is especially important when the learned dynamics model is to be used to predict over a longer horizon, resulting in compounding errors of inaccurate models. Bayesian Deep Learning approaches offer a promising alternative that allows to quantify model uncertainty explicitly, but many current approaches are difficult to scale, have high computational overhead, and poorly calibrated uncertainties. The objective for the ESR in this project will be to develop new Bayesian Deep Learning approaches, including recurrent architectures, that address these issues and are well suited for embedded control applications with their challenging constraints on computational complexity, memory, and real-time demands.

The ideal starting time is in early summer 2021. The PhD projects last for the duration of three years, and are carried out at the Neurorobotics Lab at the University of Freiburg, Germany. The PhD years include at least one longer visit – a so-called ”secondment” – between one and six months to another group in the ELO-X network, depending on the project needs and the scientific interests of the PhD fellows. The first year is mainly dedicated to studying and getting acquainted with the relevant state of the art in Bayesian Deep Learning, NMPC, and methods for scalable learning-based control. The second year focuses on development and benchmarking of optimized Bayesian Deep Learning models that can be evaluated fast, have small memory footprint, and are well calibrated to enable high-performance, robust MPC applications. In the third year, the plan is to focus on open-source software implementation, algorithm verification and integration into rapid prototyping frameworks for embedded control systems. A fourth PhD year can be added and funds are reserved for this at the University of Freiburg. The remuneration is generous and will be in line with the EC rules for Marie Curie grant holders. It consists of a salary augmented by a mobility allowance, resulting in a net monthly salary of about 1900-2300 Euro depending on family status.

Supervisors and Main Contacts

Supervising team at the University of Freiburg:

• Prof. Dr. Joschka Boedecker (head of Neurorobotics Lab)
• Gabriel Kalweit (PhD student focusing on data-efficient learning methods for control),
• Maria Huegle (PhD student focusing on Deep Learning based dynamic architectures for prediction and control).

Candidate Profile

Ideal candidates have a master degree in one of the following disciplines or a related field: computer science, control engineering, physics or applied mathematics. They should have a good background or interest in machine learning, especially Deep Learning, mathematical optimization, and programming (Python, C/C++), as well as a desire to contribute to the development of open-source software and the success of real-world experiments. Proficiency in English is a requirement. The positions adhere to the European policy of balanced ethnicity, age and gender. Both men and women are encouraged to apply.

Application

To apply, send an email to elo-x@imtek.uni-freiburg.de, attaching one single PDF file containing all contents or links (any other information within the email will not be processed). Subject of your email should be: “ELO-X PhD Application – ESR 3”.

 

Please include, in your single PDF document, the following items in this order:

 

1. A cover letter incl. statement of research interests and career goals (max. 2 pages);
2. An academic CV;
3. Contact details of at least two referees incl. phone numbers and emails;
4. Your diplomas and transcript of course work and grades;
5. Sample of technical writing (publication or thesis);
6. Proof of English language proficiency test results.

 

Please send your application as soon as possible, the call for applications closes January 17, 2021.

 

Note that your PDF will be forwarded to several people in the ELO-X institutions and that in particular all Supervisory Board members of ELO-X will have access to your application material. If you want to apply to more than one ELO-X position, please create and send separate PDFs.

Marie Curie Fellowship Eligibility Criteria in Short

To be eligible, you need to be an “early stage researcher” i.e. simultaneously fulfill the following criteria at the time of recruitment: