Collaborations in DDMA

DDMA collaborates with researchers at U Oregon in psychology and related computational science fields in the context of brain imaging. In particular, we work with the UO Neuroinformatics institute, and the related University spin-off company, Electrical Geodesics, Inc (EGI). Our collaboration to date has focused on problem solving environments software for implementing complex data analysis programs using familiar tools such as Matlab and GNU Octave. Researchers at UOregon have built tools that add functionality not found in LANL developed tools, such as the ability to run analyses in parallel and distributed environments. We are currently working to integrate these capabilities into the DDMA Information and Data Analyzer (IDA) toolkit, and they are looking at using some of IDA's capabilities for their own work.

We are building a program at present to increase collaborations in the development of analysis techniques and algorithms themselves. This work is tied to both the Neuroinformatics institute at UOregon and the Computational Science institute centered in the UOregon Computer and Information Sciences department. Preliminarly work involves image processing for the study of neural activity in the c. elegans organism in conjunction with the laboratory of Shawn Lockery in the UOregon Biology department. We are also pursuing collaborations with the behavioral psychology department, developing algorithms for video processing and data fusion of video data with eye tracking information to study reactions of the eye movement function in the brain to behavioral features in video scenes. Both of these projects are slated for work starting this summer at the LANL/CDDMA summer school.

Key personnel at UOregon: Allen Malony, Don Tucker, Shawn Lockery, Bertram Malle

Key organizations at UOregon:

• Dept. of Computer and Information Sciences
• Neuroinformatics Center
• Computational Science Institute
• Electrical Geodesics, Inc.
• Dept. of Biology
• Dept. of Psychology

LANL Collaborations on Computer Science

The DDMA collaboration builds on a long-standing collaborative relationship with the University of Oregon that has spanned nearly 15 years. Existing collaborations have been in the area of large-scale parallel program performance analysis, as manifest in the ASC-funded Tuning and Analysis Utilities (TAU) package. In the process of developing their performance analysis tools, we have collaborated with UOregon on the Program Database Toolkit package, used to perform automated source analysis and performance instrumentation. This has led to preliminary work recently towards implementing front-end tools for the Fortran 2003 standard contributed to the GNU Fortran compiler.

We have also collaborated extensively with the University of Oregon in conjunction with Lawrence Livermore, Sandia, Oak Ridge, and Argonne National Laboratories, as a core member of the DOE SciDAC Common Component Architecture effort. Our work with the University of Oregon on CCA has been in the design of component techniques for performance analysis, automated component wrapping tools based on the PDToolkit, and the CCAIN CCA-lite component framework. Our recent work with frameworks has been adopted as one of the core software technologies for the recently initiated SciDAC-2 CCA effort.