Cosmological Inference

Cosmological inference aims to extract cosmological parameters and test models from astronomical observations, a task complicated by the high dimensionality of data and model parameters. Current research heavily utilizes machine learning, employing techniques like normalizing flows, neural networks (including convolutional and graph neural networks), and Gaussian processes to emulate likelihood functions, accelerate Bayesian inference (e.g., via Hamiltonian Monte Carlo or nested sampling), and efficiently combine diverse datasets. These advancements significantly reduce computational costs and improve the accuracy and robustness of cosmological parameter estimation, enabling more sophisticated analyses of large-scale structure and other cosmological probes.