28 October 2020

The Novo Nordisk Foundation supports data science at MATH

Research grant

Professors Susanne Ditlevsen and Niels Richard Hansen from the Department of Mathematical Sciences have been granted a total of DKK 16.6 million to explore new methods for statistical inference and to identify causal relationships in large amounts of data from the biology and health areas.

Distinguished Data Science Investigators Susanne Ditlevsen and Niels Richard Hansen
Distinguished Data Science Investigators Susanne Ditlevsen and Niels Richard Hansen

The Novo Nordisk Foundation will with their new Data Science Initiative support “Development of new algorithms, methods and technologies within data science, artificial intelligence (…), data engineering, data mining, statistics, applied math, computer science, big data analytics, etc.”

This initiative puts methodological and foundational research in data science in focus, and the foundation awarded two Distinguished Data Science Investigator grants from this initiative to the two research leaders from the Department of Mathematical Sciences: Professor Susanne Ditlevsen and Professor Niels Richard Hansen, both statisticians.

The two grants will finance the employment of five PhD students and four postdocs.

Developing statistical methodology for neuroscience and ecology

Susanne Ditlevsen calls her new research project CUSP: Statistical inference for coupled stochastic processes with multiple timescales and changing environments.

The project lies at the intersection between statistics, dynamical systems and biology, with the scientific goal of developing statistical methodology for interacting stochastic processes with multiple time scales and under external forcing, with two main applications in mind, neuroscience and ecology. Susanne Ditlevsen explains:

“Modern empirical methods hugely increase the amount and type of data we collect, and the importance and relevance of complex models are increasing. The project goal is to find more principled ways of statistical inference in this type of models by hybrid methods, and for certain non-stationary classes of processes, by using co-integration analysis.

“The hypothesis is that by splitting the model into components, a large estimation problem can be split into smaller estimation problems. This is expected to lead to more robust and computationally efficient statistical inference. In particular, approximations are only implemented where necessary and the computationally intensive methods only used in the model parts where it is needed”.

For Susanne Ditlevsen’s research group two PhD students and two postdocs will be employed. Research assistant Predrag Pilipovic and postdoc Marie Levakova from the Department are already working on the project.

Clues in unstructured events

Niels Richard Hansen’s project is called CLUE: Causal Learning with Unstructured Events. He describes the project as follows:

“The fundamental question in science is the causal question "why?" Why do some develop depression? Why does prostate enlargement develop into cancer for some and not others? My ambition as a data scientist is to answer why-questions from observational data, and I believe we live in such amazing times that the data deluge will help us do so.

“The statistical proverb correlation is not causation expresses that the path to answering the why-question from observational data is fraught with difficulties. We know how to navigate the path when using structured data on a few and well-defined events, but we will go beyond that and use all available clues from unstructured event data to answer the why-question.

“My research is at the forefront of causal structure learning for asynchronous events based on structured data. My ambition is to expand into the use of unstructured data, such as medical or public records, logbooks, or patient diaries. I will develop the necessary mathematical language, the causal semantics and the learning methodologies that are currently not available. This is a foundational development that is absolutely necessary if we want to understand all the causal clues we can find in unstructured event data”.

For Niels Richard’s research group the department will employ two postdocs and three PhD students, who will all join Copenhagen Causality Lab.

The vacancies in the two research groups have been advertised and can be applied for before 15 November 2020.