Cosmological Simulation

Cosmological simulations model the universe's large-scale structure formation, aiming to accurately predict and understand its evolution from initial conditions. Current research heavily utilizes machine learning, employing neural networks (including convolutional, generative adversarial, and diffusion models) and other algorithms like neural operators and DeepONets to accelerate simulations, improve resolution (super-resolution), and efficiently infer cosmological parameters from observational data. This work is crucial for analyzing increasingly large datasets from cosmological surveys, enabling more precise constraints on fundamental cosmological parameters and a deeper understanding of galaxy formation and evolution.