Resolving cosmological tensions with diverse data, novel theories and Bayesian machine learning

Our Universe is expanding faster than we expected. The standard model of cosmology predicts a Hubble constant which differs substantially from what we measure it to be. Far from being a problem, this is exciting as it indicates the need for a new model of the Universe. Despite three years of effort however, cosmologists have been unable to resolve this cosmic conundrum.

The reasons for this community failure are threefold. First, no single theoretical solution is capable of satisfactorily resolving all discrepancies. Second, disentangling new physics from measurement error is a challenging unsolved problem. Third, our simulation and data analysis pipelines have been designed and tuned in the context of the standard model, which can bias even the most carefully designed approach.

This ambitious project proposes to resolve all three of the above and uncover and establish the next cosmological paradigm for theory and data analysis. An interlocking programme of theory, inference and observational research, undertaken by the PI, three postdocs and four PhD students over five years will aim to simultaneously resolve the tensions in both cosmological theories and data processing.

The broad aims of the project are to (a) Resolve the tensions between cosmological observations with a new standard model of the universe and next-generation numerical techniques (b) Establish likelihood-free inference at the heart of our cosmological analysis toolkit in preparation for the future onslaught of big cosmological data, and (c) Bring together a diverse set of cosmological and particle physics datasets and organise them in a coherent statistical framework.

This is an essential and substantial research effort which only an ERC starting grant can support.