A new technique combining artificial intelligence trained by supercomputer simulations and astronomy Big Data has enabled astronomers to analyze data with undreamed of speed to determine the unknown characteristics of the Universe.
In the equations governing the evolution of the Universe, there are certain unknowns which cannot be measured directly and must be inferred. To estimate these cosmological parameters, researchers produce multiple simulations of the evolution of the Universe, each with slightly different parameters, and see which best matches actual observations. But these simulations are time consuming and expensive.
A team of researchers, led by former Kavli Institute for the Physics and Mathematics of the Universe (Kavli IPMU) Project Researcher Yosuke Kobayashi (currently a Postdoctoral Research Associate at The University of Arizona), developed an emulator with artificial intelligence technology that can look at the statistical distribution of galaxies in a simulation or real data and accurately guess the parameters which lead to the observed pattern. Whereas even one simulation requires dozens of hours on a supercomputer to complete, this emulator can produce results in 1 CPU second on a laptop. The team trained the emulator on simulated data produced by "ATERUI II," the world's most powerful supercomputer dedicated to astronomy, operated by the National Astronomical Observatory of Japan (NAOJ).
Then the researchers used the emulator on real data obtained by the Sloan Digital Sky Survey for galaxy distribution. The results confirm that matter accounts for only about 30% of the energy in the Universe, the remaining 70% being the dark energy causing the accelerated expansion of the Universe. In addition, the research team succeeded in measuring with an accuracy of about 5% a cosmological parameter that represents the degree of inhomogeneity in the structure of the present Universe. This is a level of accuracy that could not be achieved with conventional analysis methods and demonstrates the usefulness of emulators in cosmology research.
"These emulators have been used in cosmology studies before, but hardly anyone has been able to take into account the numerous other effects, which would compromise cosmological parameter results using real galaxy survey data. Our emulator does and has been able to analyze real observation data," comments Kobayashi. Takahiro Nishimichi at Kyoto University, a member of the research team, adds, "Since the project was launched in 2015, we have built a simulation database, implemented machine learning, and conducted in-depth validation using simulated data before finally arriving at the analysis of data measured from the real Universe. It is a very emotional moment."
Combined with new data from ongoing and coming-up cosmological surveys with instruments such as the Prime Focus Spectrograph to be mounted on NAOJ's Subaru Telescope, this emulator will be able to tell us even more about the Universe in which we live.
These results appeared as Kobayashi et al. "Full-shape cosmology analysis of SDSS-III BOSS galaxy power spectrum using emulator-based halo model: A 5% determination of σ8" in Physical Review D on April 20, 2022.
