Peter Taylor
Observational Cosmologist
About Me
I'm currently a CCAPP fellow at the Ohio State University. Prior to this I was an NPP fellow at NASA/Caltech's Jet Propulsion Laboratory after completing my PhD at University College London's Mullard Space Science Laboratory. I use data from the largest spectroscopic and photometric surveys to test ΛCDM and fundamental physics.
Research Highlights
Posterior Combination with Normalizing Flows
I have shown how to massively accelerate the sampling of independent likelihoods using normalizing flows. Using this technique I combined DES, BOSS and CMB lensing constraints leading to some of the tightest cosmological constraints from large scale structure to date.
ML Searches for Parity-violation
Following possible detections of parity-violations in BOSS data, I developed an unsupervized machine learning approach to search for parity-violations in spectroscopic galaxy surveys. If physical in origin, this signal would point to novel physics during the epoch of inflation. My approach is free from the usual theoretical uncertanties associated with estimating the significance of the detection from simulations or theory.
Synergies of Photometric and Spectroscopic Surveys
I am the science-PI of a 3-year NASA Astrophysics Theory Program Grant to exploit the synergies between upcoming photometric and spectroscopic surveys. In this role I have shown how to combine cosmological information from photometric and spectroscopic surveys.
More Precise Weak Lensing Cosmology
I led the development of a new technique called
\(k\)-cut cosmic shear (or sometimes \(x\)-cut cosmic shear) which allows us to extract significantly more information in the face of small scale modelling uncertainties caused by baryonic feedback. This is the dominant source of theoretical uncertainty in weak lensing analyses. I demonstrated the advantages of the method using Dark Energy Survey data, and the method is now integrated into the Euclid likelihood pipeline.
Statistical Inference for Weak Lensing
I showed that modern density estimation can be used to infer cosmological parameters from full forward model realizations of the weak lensing data. By comparing this approach to the standard Gaussian likelihood approach, I demonstrated for the first time that the Gaussian likelihood approximation is unbiased. Previously, it had been conjectured that this approximation was responsible for the \(S_8\) tension.
Non-parametric Cosmology with Weak Lensing
Using data from the Canada-France-Hawaii Telescope Lensing Survey I measured the background expansion of the Universe and growth of structure free from any assumption about the underlying cosmological model. This was the first model independent measurement using weak lensing data. Both the growth and expansion were consistent with our expectation from the CMB and ΛCDM.
Searches for Self-Interacting Dark Matter
I developed a strong lensing test to search for tails of scattering dark matter particles from halo collisions. Using HST data, I initially found weak evidence for a tail of scattered particles in Abell 3827. In an updated analysis using ALMA and the VLT, we found this was due to a misidentification of multi-lensed images.
Teaching Resources
In 2022 I was invited to give a lecture on likelihoods and inference in cosmology at the Euclid Advanced School in Les Houches, France. I hope this talk may be of some use to new students in the field.