Computer Science and Mathematics Projects

The goal of this project is to research effective information visualization and visual analytics approaches for incremental machine learning based extreme-scale data analysis. This project will be aligned with two existing LDRDs (one on streaming visualization and the other on deep structured learning) and focus specifically on the effective interaction between the two approaches. The aim is to research visualization paradigms and feedback mechanisms for machine learning algorithms, as well as incremental machine learning and statistics algorithms to extract the necessary information from the data analysis pipelines. A key aspect of the work will be the scalability of the developed approaches, both in terms of the application performance and the ability of the applications to represent and convey extreme-volumes of information to the system users.

The research will be guided by example application scenarios from the NSLS II. We will collaborate with the Data acquisition and analysis team of NSLSII, to evaluate suitable data analysis and visualization paradigms for NSLSII beamlines, such as HXN, FXI and CMS. Likely target analysis pipelines are: streaming tomography, X-ray diffraction analysis, and fluorescence fitting. A preliminary experiment has been studied for diffraction data.

Outcomes

Software packages, prototype widgets or systems, paper publications on related journals and conferences

Publications

E. Papenhausen, B. Wang, M. H. Langston, M. Baskaran, T. Henretty, T. Izubuchi, A. Johnson, C. Jung, M. Lin, B. Meister, K. Mueller and R. Lethin, Polyhedral user mapping and assistant visualizer tool for the R-Stream auto-parallelizing compiler, in Proc. VISSOFT, pp. 180–184, IEEE, 2015.

M. Lin, E. Papenhausen, M. H. Langston, B. Meister, M. Baskaran, T. Izubuchi and C. Jung, Optimizing the domain wall fermion Dirac operator using the R-Stream source-to-source compiler, to appear in Proceedings of Science LATTICE2015 (2015) 022. .