The new approach comes from ETH Zurich and the ultimate purpose is with the acquisition of knowledge as to how the universe came into being and what its ultimate fate might be. To begin to answer this complex problem the mystery of dark matter and dark energy needs to be understood. Dark matter is the force that appears to pull the universe together; whereas dark energy is the force that appears to be pulling it apart. The first step, and which the project is focusing on, is understanding how much of both is out there.
Dark matter
Dark matter is a type of matter that constitutes a quarter of the total energy density of the universe. Most dark matter is probably to be composed of some as-yet undiscovered subatomic particles. In contrast to normal matter, dark matter does not interact with the electromagnetic force. This means it does not absorb, reflect or emit light, making it extremely hard to detect.
READ MORE: Hubble images reveal the distribution of dark matter
Dark matter has never been directly observed. Its probable existence explains a series of puzzling astronomical observations.
ALSO READ: Map of universe’s dark matter produced
New technology
The new technology, which is based on machine learning, has a lot of similarities with facial recognition technology. Instead of examining for faces, the technology looks for signs that might suggest the presence of dark matter and dark energy, based on maps of the universe. An example is bends the path of light rays arriving at the Earth from distant galaxies (a phenomenon called weak gravitational lensing). By using the distortion, cosmologists can work backwards and develop mass maps of the sky showing where dark matter is located.
This is where the AI comes in, according to lead researcher Alexandre Refregier: “Instead of inventing the appropriate statistical analysis ourselves, we let computers do the job.” The AI undertakes the task with improved accuracy, and at twice the speed.
The research has been published in the journal Physical Review D, with the research paper titled “Cosmological constraints with deep learning from KiDS-450 weak lensing maps.”
