Abstract
In classical cosmological analysis of large-scale structure surveys with two-point functions, the parameter measurement precision is limited by several key degeneracies within the cosmology and astrophysics sectors. For cosmic shear, clustering amplitude and matter density roughly follow the relation. In turn, is highly correlated with the intrinsic galaxy alignment amplitude . For galaxy clustering, the bias is degenerate with both and , as well as the stochasticity . Moreover, the redshift evolution of intrinsic alignment (IA) and bias can cause further parameter confusion. A tomographic two-point probe combination can partially lift these degeneracies. In this work we demonstrate that a deep-learning analysis of combined probes of weak gravitational lensing and galaxy clustering, which we call DeepLSS, can effectively break these degeneracies and yield significantly more precise constraints on , , , , , and IA redshift evolution parameter . In a simulated forecast for a stage-III survey, we find that the most significant gains are in the IA sector: the precision of is increased by approximately 8 times and is almost perfectly decorrelated from . Galaxy bias is improved by 1.5 times, stochasticity by 3 times, and the redshift evolution and by 1.6 times. Breaking these degeneracies leads to a significant gain in constraining power for and , with the figure of merit improved by 15 times. We give an intuitive explanation for the origin of this information gain using sensitivity maps. These results indicate that the fully numerical, map-based forward-modeling approach to cosmological inference with machine learning may play an important role in upcoming large-scale structure surveys. We discuss perspectives and challenges in its practical deployment for a full survey analysis.
3 More- Received 17 March 2022
- Revised 14 June 2022
- Accepted 12 July 2022
DOI:https://doi.org/10.1103/PhysRevX.12.031029
Published by the American Physical Society under the terms of the Creative Commons Attribution 4.0 International license. Further distribution of this work must maintain attribution to the author(s) and the published article’s title, journal citation, and DOI.
Published by the American Physical Society
Physics Subject Headings (PhySH)
synopsis
Machine Learning Pins Down Cosmological Parameters
Published 19 August 2022
Cosmological constraints can be improved by applying machine learning to a combination of data from two leading probes of the large-scale structure of the Universe.
See more in Physics
Popular Summary
The laws of physics governing the Universe, its composition, history, and fate, can be studied by observations of large-scale distribution of matter in the sky. By measuring the shapes and positions of millions of distant galaxies, we can understand how matter clusters together and how this distribution changes over time. However, there is a major limitation to this with conventional methods: The differences among cosmological models can be easily confused by the characteristics of the evolution of galaxies. In this paper, we present an artificial intelligence (AI) system that can distinguish between the cosmological signal and the astrophysical properties of galaxy evolution, greatly reducing the uncertainties on cosmological parameter measurement.
Compared to previous methods, which used the equivalent of “pen-and-paper” theory, the AI learns from fully numerical simulations of different cosmological models. Previous methods used very simple, “hand-designed” statistics to characterize the matter distribution, while the AI automatically detects complicated patterns and synergies in the data, which would be very hard to capture with traditional statistics. This way, the AI achieves a remarkable improvement of 15 times better precision in measuring cosmological parameters.
The AI simply exploits our capacity to simulate the Universe with remarkable precision. As the power of supercomputing clusters continues to grow, we will be able to make simulations with increasingly higher degrees of realism and to include new theories in cosmological physics. Artificial intelligence will be tremendously helpful in putting theories to the test—and now we have a way to avoid confusing them with astrophysical effects.