CCAPP Seminar - Stephen Chen (Institute for Advanced Study)

ccapp logo
May 2, 2023
12:00PM - 1:00PM
PRB 4138

Date Range
2023-05-02 12:00:00 2023-05-02 13:00:00 CCAPP Seminar - Stephen Chen (Institute for Advanced Study) Speaker: Stephen Chen (Institute for Advanced Study) "Zeldovich Control Variates: Using the Cumulant Theorem to Save You CPU Hours" N-body simulations are ubiquitous in modern cosmological analyses, producing mock data with which to test analytical models on the one hand and laying the foundation for simulations-based analyses on the other. A key difficulty with these simulations is that they are beset by cosmic variance at large scales—at a fixed resolution scaling to larger volumes becomes prohibitively expensive very quickly. In this talk I will discuss a method to cancel this cosmic variance —at close to no cost—using physical intuition about large-scale bulk flows in structure formation. For galaxy samples representative of DESI, this method produces the equivalent of multiplying the simulation volume by an order of magnitude or more on large scales. As a practical example, I will describe the application of this method to building hybrid effective field theory (HEFT) emulators. PRB 4138 America/New_York public

Speaker: Stephen Chen (Institute for Advanced Study)

"Zeldovich Control Variates: Using the Cumulant Theorem to Save You CPU Hours"

N-body simulations are ubiquitous in modern cosmological analyses, producing mock data with which to test analytical models on the one hand and laying the foundation for simulations-based analyses on the other. A key difficulty with these simulations is that they are beset by cosmic variance at large scales—at a fixed resolution scaling to larger volumes becomes prohibitively expensive very quickly. In this talk I will discuss a method to cancel this cosmic variance —at close to no cost—using physical intuition about large-scale bulk flows in structure formation. For galaxy samples representative of DESI, this method produces the equivalent of multiplying the
simulation volume by an order of magnitude or more on large scales. As a practical example, I will describe the application of this method to building hybrid effective field theory (HEFT) emulators.

Events Filters: