CORDIS
EU research results

CORDIS

English EN

Data Learning on Manifolds and Future Challenges

English EN

Big Data hands over the keys to understanding the dark universe

Big Data could hold the key to some of the most complex phenomena described in science – provided that we can make sense of its dizzying quantities of information. The DEDALE project developed algorithms enabling just that and used them to measure the amount of dark matter in the universe.

SPACE

DIGITAL ECONOMY

© Ann Stryzhekin
Did you know that the universe was filled with dark matter, subtly distorting the shapes of objects we observe? Dark matter has yet to be fully understood even by the best astronomers: they believe it’s there, but it’s invisible, so they have a hard time proving it. That’s where the big data revolution kicks in. From the recently completed Dark Energy Survey (DES) to the upcoming Square Kilometre Array (SKA) – set to become the world’s largest radio telescope – the amount of data collected by new and future missions is expected to allow for important breakthroughs. But extracting this knowledge is a serious technological challenge, one that will require next-generation data analysis methods such as the ones developed under the DEDALE project. “Traditional data analysis techniques rely on trained astrophysicists making decisions on things like which objects in an image are galaxies and which are stars, how far away these objects are, can we extract the shape of these objects given the quality of the images, etc.,” explains Dr Samuel Farrens, researcher at CEA Paris-Saclay and member of the project. “These techniques are not only outdated and non-optimal, but they can’t be used when dealing with big data. It’s like looking at the front and rear mirrors while driving a car, all at the same time: We need new methods of automatic processing when humans can no longer perform the task.” DEDALE brought together cosmology and signal processing experts, mathematicians, computer scientists, astrophysicists and industrial remote-sensing experts to tackle this issue, with a focus on problems anticipated from upcoming astrophysical surveys and remote-sensing technology. Together, they investigated the use of cutting-edge signal processing and deep-learning techniques and developed efficient software to implement them. One of the results Dr Farrens is most proud of is the production of a map showing the distribution of dark matter in the universe for the DES. It proves how innovative methods can outperform classical techniques, and is a great step towards getting tighter constraints on cosmological parameters – the Holy Grail of 21st century cosmologists. Of course, that’s just one of many outstanding results. DEDALE has notably made significant strides in Point Spread Function estimation (critical for determining galaxy shapes and one of the primary objectives of the European Space Agency’s Euclid space mission); in automated galaxy redshift estimation (how far way galaxies are from Earth); and in the development of alpha shearlet systems (useful for processing natural images). In the near future, Dr Farrens hopes that DEDALE’s impact will keep making itself felt. “In remote sensing, for instance, carriers often have strong electrical consumption constraints, and the processing power required to run all imaging, enhancement and analysing algorithms may be too big for current applications. The DEDALE project was an important step in tuning and porting versatile architectures into real-time demonstrators that can tackle multiple tasks with a single algorithm,” he explains. Now completed, DEDALE continues to live within the ongoing COSMIC project, which aims to tackle problems in the reconstruction of images faced in both astronomical and biomedical image processing. This new project could have a huge impact on neonatal medicine, as performing MRI scans on infants is extremely challenging with traditional methods.

Keywords

DEDALE, big data, dark matter, dark energy survey, data analysis, square kilometre array, SKA, remote sensing, COSMIC

Project information

Grant agreement ID: 665044

  • Start date

    1 October 2015

  • End date

    30 September 2018

Funded under:

H2020-EU.1.2.1.

  • Overall budget:

    € 2 702 397,50

  • EU contribution

    € 2 702 397,50

Coordinated by:

COMMISSARIAT A L ENERGIE ATOMIQUE ET AUX ENERGIES ALTERNATIVES