Big data sheds more light on dark matter
EU-funded researchers have helped generate the most accurate map to date of dark matter, the mysterious substance that makes up 80 % of the universe. The innovative big-data technologies they used will have a significant impact on fields as diverse as astrophysics and biomedical imaging.
© sanee #144665829, source: stock.adobe.com 2019
The EU-funded DEDALE projects notable contribution to the Dark Energy Survey, an international collaboration to investigate the nature of dark matter and dark energy, is one of several achievements demonstrating far-reaching applications for novel techniques to efficiently analyse vast volumes of data whether generated by astronomers surveying the night sky or doctors scanning the human brain.
The tools developed in DEDALE have virtually limitless applications, not restricted to a single domain. As a Future Emerging Technologies project, many of the methods tested and developed were very much ahead of the game, says project communications lead Samuel Farrens at the French Alternative Energies and Atomic Energy Commission (CEA).
Creation of the map showing the distribution of dark matter in the universe exemplifies how innovative data-science techniques can outperform classical methods. It marks a milestone on the road towards attaining tighter constraints on cosmological parameters, which is the primary goal of modern cosmologists trying to determine the origins and evolution of the universe.
According to Farrens: Traditional data-analysis techniques rely on astrophysicists making assumptions about questions such as which objects in an image are galaxies and which are stars? How far away are these objects? Can we determine the shape of these objects given the quality of the images? In the era of big data, this is no longer feasible. Advances in mathematics and computer science, in particular fields such as deep learning, have led to an abundance of new tools that can be used to tackle these problems.
Automated and robust data analytics
The DEDALE team developed and implemented cutting-edge signal processing and data-analysis techniques, encompassing technologies such as machine learning, cloud computing, convex optimisation and sophisticated mathematical decompositions.
The projects open source tools enable scientists to harness big data in an automated, robust and efficient way. Thus, they can address challenges such as removing the blur introduced into galaxy images by telescopes or atmospheric disturbances, determining the distances of galaxies from the light they emit, and as in the map for the Dark Energy Survey measuring the shapes of galaxies in order to infer the amount of dark matter.
Big data is a tricky problem with a constantly evolving definition. The Dark Energy Survey gathered up to 2.5 terabytes of data for each night of observation. Over a five-year period, this adds up to what most people would agree is big data. On the other hand, the upcoming Square Kilometre Array, the worlds largest radio telescope which will start collecting data from the mid-2020s, estimates a stream of up to three terabytes every second, Farrens explains.
He expects DEDALE techniques to benefit researchers working with data from the Square Kilometre Array and on future surveys such as the European Space Agencys Euclid mission which aims to measure the acceleration of the universe to further the search for dark matter and dark energy.
From outer space to inner brain
A follow-up project, COSMIC (Compressed Sensing for Magnetic Resonance Imaging & Cosmology) is refining the techniques and developing an open source software package called PySAP. This incorporates a modular optimisation toolbox and plugin framework to enable the development of new applications for astrophysics or other fields, notably biomedical research.
For example, PySAP tools can be applied to magnetic-resonance images in combination with sophisticated sampling schemes in order to dramatically reduce the time required to scan a human brain.
There has been significant interest in our work from the biomedical imaging community for applications such as magnetic-resonance imaging and electron tomography, which require advanced image reconstruction tools to achieve high-resolution images in a short period of time, Farrens says.
DEDALE researchers are currently exploring electron-tomography applications with experts in Grenoble. They have also submitted a proposal to collaborate with a childrens hospital in Geneva that could result in significant advances in the detection and understanding of neurological disorders in new-born babies.
The potential health and social benefits from these initiatives are considerable, concludes Farrens.
© L. El Gueddari, 2019