How to predict accurate cosmological statistics in any model of gravity

This project is my MPhys dissertation, which I am currently completing under the supervision of Dr Benjamin Giblin and Prof Alkistis Pourtsidou. My work focuses on selecting the most optimal parameters to run computationally expensive cosmological simulations. I am experimenting with multidimensional sampling techniques, specifically Latin Hypercube Sampling (LHS), to optimize the distribution of training and test datasets, ensuring a large number of emulator predictions achieve a margin of error within 1%. The project also involves implementing Bayesian optimization to identify high-accuracy regions within the parameter space, enabling more efficient use of computational resources for cosmological simulations. This work combines advanced machine learning techniques with statistical modeling to address complex problems in computational cosmology.