Speaker
Description
With the advent of stage-IV spectroscopic galaxy surveys, modern cosmology has entered an era of unprecedented precision. While perturbation theory approaches remain the state of the art for large-scale structure (LSS) analyses, their applicability is restricted to the mildly non-linear regime, meaning that valuable data is being discarded. Simulation-based methods offer an alternative to push our analyses to smaller scales. However, the computational cost of running high-fidelity simulations presents a major bottleneck. In recent years, machine-learning frameworks have been developed to emulate the output of N-body codes. Building on these efforts, we present a convolutional neural network (CNN) that leverages the Evolution Mapping framework. This approach exploits cosmological parameter degeneracies, enabling the CNN to emulate N-body simulations across a broad range of cosmological scenarios.
| Other topic / keywords: | Simulations, emulation |
|---|