Code & data
Public software and simulation data from my research. More code is at github.com/j-dr.
Code
-
Differentiable cosmology likelihoods in JAX, built on fast surrogate models. Because it is differentiable, it works with gradient-based samplers. It is the pipeline behind our lensing counterterm and two-loop cosmic shear analyses.
-
An emulator for hybrid effective field theory and matter power spectra, trained on the Aemulus ν simulations. It gives fast, accurate predictions for galaxy clustering and lensing, including massive neutrinos. Formerly known as anzu.
-
A Python implementation of ADDGALS, which adds realistic galaxies (positions, colors and magnitudes) to large dark-matter simulations to produce mock sky catalogs. It was used to build the Buzzard catalogs.
Data
-
Simulated galaxy surveys covering thousands of square degrees, with billions of galaxies, lensing and photometry. They were built to test and validate Dark Energy Survey cosmology analyses, and are publicly available through NOIRLab's Astro Data Lab.
-
Suites of N-body simulations that span many cosmological models, designed for training emulators of galaxy clustering and lensing. Aemulus ν adds massive neutrinos and varies the dark energy equation of state.