PhD student in Scientific Computing focusing on Machine Learning for Scalable Parameter Inference

Uppsala University, Department of Information Technology

Uppsala University is a comprehensive research-intensive university with a strong international standing. Our ultimate goal is to conduct education and research of the highest quality and relevance to make a long-term difference in society. Our most important assets are all the individuals whose curiosity and dedication make Uppsala University one of Sweden’s most exciting workplaces. Uppsala University has over 54,000 students, more than 7,500 employees and a turnover of around SEK 8 billion.

The Department of Information Technology has a leading position in research and all levels of higher education. Today the department has 280 employees, including 120 academic staff and 110 full-time PhD students. The Department comprises research and education in a spectrum of areas within Computer Science, Information Technology and Scientific Computing. More than 4000 students take one or several courses offered by the Department each year.

The position is hosted by the Division of Scientific Computing within the Department of Information Technology. As one of the world’s largest focused research environments in Scientific Computing the research and education has a unique breadth, with large activities in classical scientific computing areas such as mathematical modeling, development and analysis of algorithms, scientific software development and high-performance computing. The division is currently in an expansive phase in new emerging areas such as cloud and fog computing, data science, and artificial intelligence, where it plays key roles in several new strategic initiatives at the University. The division currently hosts 20 PhD students, with more than 80 doctorates awarded. Several PhD alumni from the division are successful practitioners in the field of scientific computing and related areas, in industry as well as in academia.

The position is also part of the Science for Life Laboratory (SciLifeLab) network and offers a rich and highly interdisciplinary research environment. SciLifeLab is a leading institution and national research infrastructure with a mandate to enable cutting-edge life sciences research in Sweden, foster international collaborations, and attract and retain knowledge and talent. Our research group specializes in developing theory, methods and software for intelligent scientific experiments. We have a wide network of collaborators, and there will be opportunities to work together with excellent researchers within Sweden and abroad.

Project description: The likelihood-free parameter inference problem involves fitting the parameters of a simulation model to observed data, when the corresponding likelihood function is unavailable. We consider the specific setting consisting of stochastic simulation models, with data being observed in the form of a time series.

Available methods to solve the parameter inference problem in this setting typically involve a careful selection or sampling of parameter combinations to simulate and means to compare simulated time series to observed time series data. The two components of sampling and comparison of simulated/observed data are typically iterative in nature and may render the parameter inference process slow, particularly for large-scale problems involving a high number (>a few tens) of parameters to infer.

In recent times, machine learning has accelerated various parts of the parameter inference pipeline, e.g., the use of neural networks as summary statistics of stochastic time series, and neural density estimators as surrogate models of the simulator. The goal of this project is to develop machine learning methods to enable scalable, efficient and accurate parameter inference of stochastic simulation models (gene regulatory networks in particular). Some research questions to  explore include, how do we efficiently select/sample parameters that are more likely to lead to the true parameters, how do we accurately compare high-dimensional time series of varying resolutions, and how do we efficiently estimate a posterior distribution using the sampling information while minimizing the number of simulations required.

Duties:
The duties of a PhD student are primarily directed at their own research education, which lasts four years. The work may also involve, to a limited extent (ca 20%) other departmental duties, such as teaching undergraduate courses and administrative tasks – in which case the position may be extended to a maximum of five years.

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 requirements for this position include basic knowledge of, and interest in machine learning, and proficiency in programming (e.g., in Python/Julia). Extra merits with equal weights include knowledge and experience in numerical optimization, Bayesian methods and deep learning, bioinformatics and/or computational biology, and best practices in software engineering.

Application: Your application should include

• a personal letter describing your background, relevant qualification, research interests and how you see that your background would contribute to the specific project.
• Curriculum Vitae,
• degree documents (for any completed degree) and transcript of records (including completed courses and grades) to support that you’re qualified for admission to education at the PhD level,
• list of publications (if any),
• copy of your master thesis, or a detailed description of the ongoing thesis work, including a plan for its completion, if not fully complete at the end of the application period,
• list of reference persons with contact details,
• any other supporting documentation demonstrating your suitability for the position.

It is not a strict requirement for the relevant pre-requisite degree to be completed when the application is made. All applicants should state when they would be able to fulfill all requirements and be ready to assume the position.