Forecasting and Optimization in Modern Cosmology

Doctoral Thesis

2010

Permanent link to this Item
Authors
Supervisors
Journal Title
Link to Journal
Journal ISSN
Volume Title
Publisher
Publisher

University of Cape Town

License
Series
Abstract
Cosmology is emerging into a new and exciting period thanks to a wealth of ongoing and planned massive surveys which will deliver exponentially growing volumes of data over the next two decades. As a result of this rapid growth, which exhibits erce competition between di erent surveys due to the spiralling costs, forecasting and optimization have become critical to help best use and bene t from this new boon. In this thesis various aspects of forecasting and optimization are explored, with particular emphasis on, but not limited to, cosmology. We introduce a new optimization algorithm which signi cantly outperforms all standard algorithms, especially in higher dimensions where the improvement is remarkable. The new algorithm, Hybrid-MTM, should provide a powerful new tool in addressing high-dimensional optimization problems. We then forecast the prospects for detecting dynamics in tracking dark energy models. We show that Big Bang Nucleosynthesis and Cosmic Microwave Background constraints in these models are extremely di cult to match with existing data. As a result it is unlikely that a detectable deviations from the cosmological constant for these models is possible before the Stage-IV DETF experiments, which will only come on-line post-2015. Finally we present new results on Fisher matrix forecasts for cosmology produced using the Fisher4Cast code. Fisher4Cast allows novel insights into the nature of how information is gained from cosmological experiments and the interplay between the measurements of Hubble, distance and growth in constraining cosmological parameters. In the nal chapter we provide a detailed overview of the code structure in Fisher4Cast and its Graphical User Interface together with its unique features including a LATEXreporting module which breaks new ground in the automated generation of publication-quality scienti c research.
Description

Reference:

Collections