Network based systems

Neural network analyzes gravitational waves in real time – sciencedaily

Black holes are one of the biggest mysteries in our Universe – for example, a black hole with the mass of our Sun has a radius of only 3 kilometers. Black holes orbiting each other emit gravitational radiation – oscillations of space and time predicted by Albert Einstein in 1916. This makes the orbit faster and narrower, and eventually, the black holes merge into a last burst of radiance. These gravitational waves travel through the Universe at the speed of light and are detected by observatories in the United States (LIGO) and Italy (Virgo). Scientists compare data collected by observatories with theoretical predictions to estimate the properties of the source, including the size of black holes and their rotational speed. Currently, this procedure takes at least hours, often months.

An interdisciplinary team of researchers from the Max Planck Institute for Intelligent Systems (MPI-IS) in Tübingen and the Max Planck Institute for Gravitational Physics (Albert Einstein Institute / AEI) in Potsdam are using cutting-edge machine learning methods to accelerate this process. They developed an algorithm using a deep neural network, a complex computer code constructed from a sequence of simpler operations, inspired by the human brain. In a few seconds, the system deduces all the properties of the binary source of the black hole. The results of their research have been published in the flagship journal of Physics, Physical examination letters.

“Our method can make very precise statements within seconds about the size and mass of the two black holes that generated the gravitational waves when they merged. How fast are black holes spinning, how far apart are they of the Earth and from which direction is the gravitational wave coming? We can deduce all this from the observed data and even make statements about the precision of this calculation “, explains Maximilian Dax, first author of the study Real-time gravitational wave science with posterior neural estimation and Ph.D. student in the Empirical Inference Department at MPI-IS.

The researchers trained the neural network with numerous simulations – gravitational wave signals predicted for hypothetical binary black hole systems combined with noise from detectors. In this way, the network learns the correlations between the measured gravitational wave data and the parameters characterizing the underlying black hole system. It takes ten days for the algorithm called DINGO (the abbreviation means Dthick INreference for gravitational wave ohobservations) to learn. Then it’s ready to go: the array derives size, spins, and all other parameters describing black holes from newly observed gravitational wave data in just a few seconds. High-precision analysis decodes ripples in space-time in near real time – something never before done with such speed and precision. The researchers are convinced that improving the performance of the neural network as well as its ability to better manage noise fluctuations in detectors will make this method a very useful tool for future gravitational wave observations.

“The more we explore space through increasingly sensitive detectors, the more gravitational wave signals are detected. Fast methods like ours are essential to analyze all this data in a reasonable amount of time,” says Stephen Green, Principal Scientist in the Department of Astrophysical and Cosmological Relativity at AEI. “DINGO has the advantage that once trained, it can analyze new events very quickly. Importantly, it also provides detailed parameter uncertainty estimates, which were difficult to produce in the past using machine learning methods. So far, researchers from the LIGO and Virgo collaborations have used very time-consuming computational algorithms to analyze the data. They need millions of new gravitational waveform simulations for the interpretation of each measurement, which leads to computation times of several hours to several months. known as “damped inference”.

The method is promising for more complex gravitational wave signals describing bit patterns of black holes, the use of which in current algorithms makes analyzes very long, and for binary neutron stars. While the collision of black holes releases energy exclusively in the form of gravitational waves, the fusion of neutron stars also emits radiation in the electromagnetic spectrum. They are therefore also visible by telescopes which must be pointed towards the respective region of the sky as quickly as possible in order to observe the event. To do this, it is necessary to determine very quickly where the gravitational wave is coming from, as facilitated by the new method of machine learning. In the future, this information could be used to point telescopes in time to observe electromagnetic signals from collisions of neutron stars and a neutron star with a black hole.

Alessandra Buonanno, Director of AEI, and Bernhard Schölkopf, Director of MPI-IS, are delighted at the prospect of taking their successful collaboration to the next level. Buonanno expects that “in the future, these approaches will also allow much more realistic processing of detector noise and gravitational signals than is possible today using standard techniques,” and Schölkopf adds that such “simulation-based inference using machine learning could be transformative in many areas of science where we need to infer a complex model from noisy observations. “

Source of the story:

Materials provided by Max Planck Institute for Intelligent Systems. Note: Content can be changed for style and length.

Source link