Three PhD student positions in optimization for data-driven modeling and machine learning - Hiring in process/Finished, not possible to apply

At the Division of Systems and Control we develop methodology for and applications of automatic control, system identification  and machine learning. An important goal is to develop mathematical models that can capture real-world dynamical phenomena and their environments allowing machines and humans to act efficiently in  the world around us. Optimization methods are of central importance since they are widely applied in control, system identification, and machine learning. Quantification of uncertainty is an important aspect in modeling of complex dynamical systems and phenomena.

The Division of Systems and Control enjoys a wide network of strong international collaborators all around the world, for example at the University of Cambridge, University of Oxford, Imperial College, University of British Columbia, University of Sydney, University of Newcastle and Aalto University. We strive for all PhD students to get a solid international experience during their PhD.

The research projects for the advertised positions will be within the areas of optimization and mathematical modeling of dynamical systems, both from data and underlying physical principles. They feature for instance machine learning, Bayesian estimation, and optimization based on models in the form of partial differential equations.

As an applicant you are not required to specify a specific research topic in your application, but you are  most welcome to do so if you wish. The precise research topic of each PhD student will be decided in a dialog between the student and the supervisor after a successful appointment. The three problem formulations that are most relevant for this opening are the following:

1. Optimization of deep brain stimulation:

Deep brain stimulation (DBS) is an invasive treatment for predominantly neurological and mental conditions. DBS utilizes continuous stimulation of a carefully selected target area in the brain with electrical pulses through an implanted electrode. In principle, DBS establishes direct contact between a population of neurons within the stimulation target and an implanted in the patient’s body computer that controls the neural population in order to alleviate the symptoms of the disease. The DBS signal is selected so that the target is stimulated to achieve the best therapeutic result and the stimuli spill outside of the target is minimized. Therefore, the DBS signal tuning can be formulated as an optimization problem and solved numerically by making use of an individualized mathematical model that describes the involved part of the brain, the DBS electrode, and the stimulation signal.

This project is funded by the Swedish Research Council and carried out in cooperation with the Uppsala University Hospital within an international consortium comprising  Charité Berlin, University of Amsterdam, University of Luxembourg and others.

2. Data-driven epidemiological modeling:

The ongoing COVID-19 pandemics has highlighted issues with our preparedness to handle epidemics on both regional as well as on national level. A particular issue concerns our ability to effectively analyze data and develop useful and accurate prediction models. The Coronavirus is a zoonotic virus that has spread from animal to human, and continuous surveillance of such diseases is another important aspect. Also in this area is data-driven models are used for decision support and situation awareness. Interesting challenges include epidemiological modeling under large uncertainties and with data of low information content. In the area of infectious diseases with an impact on society, estimates of uncertainties and confidence bounds are particularly important goals.

This project contains two sub-projects and a specific final application: (1) effective and informative simulation software, (2) Bayesian modeling of disease spread under large uncertainties, and (3) a case study of Salmonella Dublin. The software part includes contributions of computational models within SimInf (www.siminf.org), in particular with a focus on effective propagation of uncertainties. Bayesian modeling is a consistent way of informing a mathematical model with data and in such a way that uncertainties propagates naturally and may aid in estimating risks of different outcomes. The case study of Salmonella, finally, is a specific dataset collected for an endemic/static state. The challenge here is to produce a specific model which can provide for decision support at the national level. For this project, the candidate could be placed at either the Division of Systems and Control or at the Division of Scientific Computing, depending on the candidate’s interest.

The project is financed by Formas and conducted together with the Swedish Veterinary Institute, SVA.

3. Improved optimization using machine learning

Optimization powers most of machine learning, deep learning not the least. But what happens if you turn this around, and instead use machine learning to power optimization? This question challenges long-held beliefs in optimization and is at the forefront of so-called data-driven optimization—an emerging research field that aims to develop the next generation of optimization methods by combining different aspects of modern machine learning. As the following three examples show, the context determines which aspect is the most appropriate: (1) for stochastic optimization, a natural fit is probabilistic numerics, wherein conventional methods are reframed as probabilistic ones, (2) for constrained optimization graph neural networks, which model relations between variables as a graph, are well-suited, (3) if the goal is to embed the optimization solver in a larger system that you want to train end-to-end, then a good choice is differentiable programming, where the optimizer itself is made automatically differentiable. Since the field is still in its infancy, there are many exciting research problems to choose from, and the PhD student will have a large freedom in steering towards the ones he or she finds the most interesting.

This position is funded by Wallenberg AI, Autonomous Systems and Software Program (WASP), Sweden’s largest individual research program ever and a major national initiative for strategically motivated basic research, education and faculty recruitment. The program addresses research on artificial intelligence and autonomous systems acting in collaboration with humans, adapting to their environment through sensors, information and knowledge, and forming intelligent systems-of-systems. The vision of WASP is excellent research and competence in artificial intelligence, autonomous systems and software for the benefit of Swedish industry. Read more: https://wasp-sweden.org/

The PhD student will take part in WASP’s graduate school, which is dedicated to providing the skills needed to analyze, develop, and contribute to the interdisciplinary area of artificial intelligence, autonomous systems and software. Through an ambitious program with research visits, partner universities, and visiting lecturers, the graduate school actively supports forming a strong multi-disciplinary and international professional network between PhD students, researchers and industry. Read more: https://wasp-sweden.org/graduate-school/

Duties/Project description: The position is for a maximum of five years and includes departmental duties at a level of at most 20% (typically teaching).

Requirements: A PhD position at the Division requires a Master of Science or equivalent in a field that is relevant to the topic of the PhD thesis, good communication skills and excellent study results, as well as sufficient proficiency in oral and written English.

Additional qualifications: Experience in optimization, machine learning (or computational statistics), computational modeling or dynamical systems is valued.

The application should include a statement (at most 2 pages) of the applicant’s motivation for applying for this position, including the candidate’s qualifications and research interests and evidence of self-motivation and constructive teamwork. The application should also include a CV; degrees and grades (translated to English or Swedish); the Master’s thesis (or a draft thereof, and/or some other self-produced technical or scientific text), publications, and other relevant documents. References with contact information and up to two letters of recommendation may be provided. Applications may be submitted by candidates that have not fully completed the Master of Science degree (or equivalent), however all applicants should state the earliest possible starting date of employment.

Rules governing PhD students are set out in the Higher Education Ordinance chapter 5, §§ 1-7 and in Uppsala University's rules and guidelines.

Salary: According to local agreement for PhD students.
 
Starting date: As soon as possible or as otherwise agreed.

Type of employment: Temporary position according to the Higher Education Ordinance chapter 5 § 7.

Scope of employment: 100 %

For further information about the position, see http://www.it.uu.se/ (the department) or contact: Prof. Alexander Medvedev (alexander.medvedev@it.uu.se), Prof. Stefan Engblom (stefan.engblom@it.uu.se), or Dr Jens Sjölund (jens.sjolund@it.uu.se).

Please submit your application by 19 April 2021, UFV-PA 2021/1044.

Are you considering moving to Sweden to work at Uppsala University? Find out more about what it´s like to work and live in Sweden.