Speaker: Jamie Donald-McCann

Title:  Cosmic Emulation with Simple Neural Networks

Abstract: Future galaxy surveys like DESI and Euclid will map the positions of 100s of millions of galaxies, and will allow us to probe the structure in the universe with unprecedented accuracy. To obtain the maximum possible information from these surveys the accuracy of theoretical predictions for galaxy clustering on non-linear scales needs to be improved. These non-linear scales offer increased statistical power, and are most sensitive to signatures of modified gravity and massive neutrinos. In this talk I will discuss emulation as a method for producing accurate predictions of galaxy clustering on non-linear scales, I will cover some general approaches to emulation, and will outline some of the key steps to developing an emulator for clustering statistics. I will also highlight some initial results from an investigation into using simple neural networks as emulators, and constraints on HOD and cosmological parameters that can be achieved using these emulators.

Meeting Recording:
https://port-ac-uk.zoom.us/rec/share/33MbONYzV4AshQwekcb6pgO3tYAeHTt9uqXaUMP2RL_Im4pnJ2VtGnL1g1PlNiu9.s0mmA7ECSdW4bbNz

Slides:

https://drive.google.com/drive/folders/1lF3tJHNarrcqE-d1BkBM-4dXixbfGa3J?usp=sharing