Abstract :
This talk will present my work on computational and statistical methods for extracting cosmological information from the large-scale structures (LSS) of the Universe, with particular emphasis on the search for primordial non-Gaussianity (PNG), a key observational signature for distinguishing between models of inflation. I will discuss how higher-order statistics, and in particular the bispectrum, can access information that is not captured by the power spectrum, making them powerful tools for probing primordial physics. The talk will cover my previous work on bispectrum forecasts for current and future surveys, including applications to Euclid, as well as the development of efficient pipelines based on modal decompositions and simulation-based approaches. I will then describe more recent work with cosmological simulation suites, where forward-modelling and likelihood-free inference techniques are used to extract cosmological information from LSS datasets. A particular focus will be on recent developments in field-level inference, where Bayesian forward models aim to constrain PNG directly from the density field rather than through compressed summary statistics. I will conclude by outlining future directions, including applications to CMB–LSS cross-correlations, 21 cm cosmology, machine learning, and next-generation cosmological inference.
Short Bio: Dionysis Karagiannis is a postdoctoral researcher at the University of Groningen, working on computational cosmology and methods for extracting information from large-scale structures (LSS), cosmological simulations, and 21 cm intensity-mapping data. His research aims to use current and future cosmological surveys to test models of the early Universe, with a particular focus on primordial non-Gaussianity and higher-order statistics. He received my PhD in Physics from the University of Padova, where he worked on the modelling, prediction, and estimation of the LSS bispectrum. He has since held research positions at the University of Padova, the University of the Western Cape, Queen Mary University of London, and the University of Ferrara, working across optical and radio cosmology, multi-tracer analyses, HI intensity mapping, and large cosmological simulation suites. He is an active member of several international collaborations, including Euclid, SKA, and HIRAX. His recent work develops forward-modelling, likelihood-free, and Bayesian field-level approaches to extract cosmological information from current and future LSS surveys.
