AI for Cosmology

This chapter describes the usage of ML in cosmology. The Standard Model of cosmology is summarized. In order to infer the global shape of the universe, Bayesian methods are applied to experimental data like the total energy density of the universe – originating from ordinary and dark matter, along with dark energy. This analysis suggests odds of the order of 50:1 or 200:1 in favor of a flat compared to a closed or open universe, respectively, and probability between 67 and 98 percent that the universe is spatially infinite. Applied to neutral hydrogen’s characteristic radio frequency, Bayesian methods are employed to distinguish new physics including the interaction of hydrogen with dark matter from instrumental effects. The expansion of the universe is studied using ML to quantify the redshift of light from far distant galaxies. Complementing size by structural information, applying ML to study weak gravitational lensing is our best bet for measuring the statistics of cosmic structure. A recently established probe for monitoring cosmological phenomena are gravitational waves, characterized by incredibly small changes in distances. Thus, differentiating gravitational wave detections from possible noise artifacts such as seismic activity and the complexity in the detectors is essential and another important task for ML.