PIRSA:20010089

Reconstruction of Cosmological Fields in Forward Model Framework - Galaxy Clustering and Intensity Mapping

Modi, C. (2020). Reconstruction of Cosmological Fields in Forward Model Framework - Galaxy Clustering and Intensity Mapping. Perimeter Institute. https://pirsa.org/20010089

Modi, Chirag. Reconstruction of Cosmological Fields in Forward Model Framework - Galaxy Clustering and Intensity Mapping. Perimeter Institute, Jan. 21, 2020, https://pirsa.org/20010089 

          @misc{ pirsa_PIRSA:20010089,
            doi = {10.48660/20010089},
            url = {https://pirsa.org/20010089},
            author = {Modi, Chirag},
            keywords = {Cosmology},
            language = {en},
            title = {Reconstruction of Cosmological Fields in Forward Model Framework - Galaxy Clustering and Intensity Mapping},
            publisher = {Perimeter Institute},
            year = {2020},
            month = {jan},
            note = {PIRSA:20010089 see, \url{https://pirsa.org}}
          }

Chirag Modi

University of California, Berkeley

Talk number
PIRSA:20010089
Collection
Talk Type
Subject
Abstract

In this talk, I will outline the forward model approach to reconstruct cosmological fields in a Bayesian framework. I will focus on two examples - galaxy clustering and neutral hydrogen intensity mapping.

In galaxy clustering example, I will use the observations of galaxy surveys to reconstruct the initial Lagrangian field. Here, we develop a novel framework with neural networks to forward model halo masses and positions and demonstrate that our method outperforms standard reconstruction in both real and redshift space. This reconstructed initial field has enhanced signal for baryon acoustic oscillations and can enhance science returns for surveys like DESI.

For neutral hydrogen surveys, we lose over 50% of the modes at high redshifts due to foregrounds and it severely hampers their feasibility for cosmological analysis. With a novel bias framework for the forward model, I will show that we are able to reconstruct over 90% of these modes and this recovers cross-correlations with photo-z surveys like LSST and tracers like CMB lensing.

Lastly, I will briefly touch upon assumptions made in this reconstruction framework regarding noise models and likelihood. I will discuss preliminary ways to improve upon them using deep learning.