Young scientists to be trained to analyse Big Data

The University of Groningen will lead a European consortium that will train 14 PhD students to analyse large databases using state-of-the-art computer science. These students will focus on galaxy evolution, but their skills could just as well be applied to image recognition in surveillance or other Big Data problems. The training programme, called Sundial,will be funded by a EUR 3.7 million grant from the EU Marie Skłodowska-Curie Innovative Training Networks programme.

The project leader is University of Groningen Professor of Astronomy Reynier Peletier. ‘What is unique about this proposal is that the students will all be trained in both astronomy and computer science.’ The nine academic partners come from both fields. The project also includes five partners from industry, one of whom is IBM Zürich. ‘They are mainly interested in machine learning, but also in visualization’, Peletier explains.

The training programme revolves around an astronomical question: how do galaxies evolve? To answer this question, astronomers study data from large surveys. Images from these surveys contain truly ‘astronomical’ numbers of galaxies, which can be extremely faint. ‘You can barely see them in these pictures’, says Peletier, who is involved in several surveys focusing on the Fornax Cluster in the Southern skies. It is not feasible to visually identify these. One of the main aims of the project, therefore, is to write software that will automatically identify and classify galaxies in the images.

The 14 students will start as one cohort in June 2017, and will be trained together and at their host institutions. Furthermore, each student will spend six months with one of the academic partners, or three months with an academic partner and three months with a partner from industry. ‘An important part of the training is to teach students the language of both astronomy and computer science’, says Peletier.

As part of the training programme, the students will learn generic skills that can be used in both academia and the private sector. Analysing Big Data is an important issue in modern society. One example is the automated analysis of images acquired by surveillance cameras. But there are many other examples where computer science is needed to extract information from large databases.

The University of Groningen will host four of the 14 PhD students. Three will work primarily at the Johann Bernoulli Institute for Mathematics and Computer Science and one at the Kapteyn Astronomical Institute. ‘Overall, the distribution between computer science and astronomy is about even’, says Peletier. The network will start advertising for students at the end of 2016.

List of academic partners (number of PhD students):

University of Groningen, the Netherlands (4); University of Birmingham, UK (2); Heidelberg University, Germany (1); University of Oulu, Finland (1); ESIEE, Paris, France (1); University of Naples Federico II, Italy (1); Osservatorio Capodimonte, INAF, Naples, Italy (1); Instituto de Astrofisica de Canarias, Spain (1). Ghent University, Belgium (2)

Companies:

IBM (Zürich, Switzerland) Cloud and computing infrastructure, and cognitive computing and computational sciences.

TARGET Holding (Groningen, the Netherlands) Big Data systems for business applications

ADCIS (Caen, France) Imaging applications

VICOMTECH (San Sebastian, Spain) computer vision, computer graphics and interaction

CLEVER-FRANKE (Utrecht, the Netherlands) data visualization, design and development