Cosmic Microwave Background Analysis Tools

Forthcoming astrophysical data sets will be so large that novel analysis techniques must be developed if the data is to be properly exploited. We propose to apply state-of-the-art information science techniques and tools to facilitate the analysis and visualization of large data sets in astrophysics, particularly in cosmology within the area of cosmic microwave background anisotropy studies. This research field is widely considered to represent the future of cosmology, and both NASA and ESA are devoting major resources towards generating sensitive all-sky maps of the cosmic microwave background. However present techniques will be entirely inadequate for the necessary data analysis even on the largest available or projected supercomputers. We adopt a comprehensive approach by addressing all aspects of the analysis and visualization problem, and by assembling a multi- disciplinary team of computer scientists, statisticians, and CMB theory data and analysis experts.

• We will implement state-of-the-art computer science serial and parallel algorithms to optimize the analysis programs that are presently used to compare data to models. Using existing public-domain packages (LAPACK, ScaLAPACK), the codes will be performance portable across high performance workstations, shared memory parallel machines, and distributed memory parallel machines.
  
• An aggressive compression of the data, using e.g. spherical wavelet transformation, and a proposed new method to extract the scientific information from small subsections of the original data set, will allow desktop workstations to perform meaningful analysis on data sets that would otherwise be available only to massively parallel supercomputers. We will develop the analysis technique and quantify the relation between the level of compression and information loss.
  
• There are presently no systematic, efficient techniques to test whether a data set is Gaussian or non-Gaussian distributed. The problem is crucial for CMB anisotropy data sets, as well as to other data sets in astrophysics. We will complete the development, programming and distribution of a novel tool, which is extremely efficient computationally, that searches and quantifies the nature of the data-sets probability distribution function.
  
• Visualizing images of the CMB anisotropy sky is an inherent part of the data analysis process. Spherical wavelets will be adapted to the visualization of CMB anisotropy images, and to their denoising/compression process. We will extend the standard Wiener filtering/denoising technique to handle the expected large data sets.

The computational tools that we develop will be based on available scientific programming environments such as IDL or MATLAB where expedient, and will be distributed to the scientific community. The final products of this effort will be documented, as a thoroughly tested, optimized set of publicly-available tools. We expect that these tools will be an integral part of the analysis of future generation CMB anisotropy experiments, including the satellite missions, and that they will be useful in other fields of astrophysics research.

The collaborators on this work come from various disciplines.