Correcting Redshift Distortions with Artificial Intelligence

Spirov, Ivan (2024) Correcting Redshift Distortions with Artificial Intelligence. Bachelor's Thesis, Astronomy.

Preview	Text ThesisIvanSpirovCorrectingRedshiftDistortions-FINAL.pdf Download (13MB) | Preview
	Text Embargo Spirov.pdf Restricted to Registered users only Download (147kB)

Abstract

Aim: The cosmological redshift of distant celestial objects is pivotal in modern astronomy, facilitating our understanding of the universe’s structure and evolution, and enabling measurements over cosmic distances. However, this technique is susceptible to distortions, particularly pronounced in dense cosmic structures like galaxy clusters and filaments, leading to inaccuracies in distance calculations. These distortions arise from the peculiar velocities of objects relative to the expanding universe. Methods: This thesis proposes a solution utilizing Artificial Intelligence (AI), trained on simulated data from the Illustris 3 Dark simulation. The neural network is designed to correct these distortions by learning patterns in the redshift data and accurately reconstructing the true spatial distribution of cosmic structures. The network demonstrates significant improvement in correcting distortions. This is observed in both individual subhalos and full density plots, generated via the Delaunay Tessellation Field Estimator (DTFE) algorithm. Results: The project showcases the effectiveness of the neural network in restoring the complex web of cosmic structures from distorted redshift observations, highlighting its ability to enhance the fidelity of cosmological datasets. The model achieves an average accuracy of over 70% on testing samples, successfully recovering overdensities and infall regions that are crucial for understanding large-scale cosmic structure.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Weijgaert, M.A.M. van de
Degree programme:	Astronomy
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	19 Jul 2024 11:41
Last Modified:	15 Jul 2025 13:38
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/33540

Actions (login required)

View Item