PYTHIA-II: A Knowledge/Data Base System for Testing and Recommending Scientific
Elias N. Houstis, Vassilios S. Verykios, Ann C. Caitlin, Naren Ramakrishnan, John R. Rice · Purdue e-Pubs (Purdue University System) · 1999
Very often scientists are faced with the task of locating appropriate solution software for their problems and then selecting from among many alternatives.Issues related to how someone specifies problems, extracts content information, builds knowledge bases, infers answers, and identifies software resources are crucial to any scientific computing development today.In [Houstis et a1.1991] we had proposed an approach for dealing with these issues by "processing" performance data obtained from "testing" software.Reliable testing requires identification of benchmarks that "densely" cover many of the application domain "features", systematic testing procedures and automatic ways to collect and analyze the results of this process.Testing constitutes a significant investment of effort and expertise that cannot be duplicated easily by an average scientific or engineering group.In this paper, we present the architecture and implementation of a knowledge/data base system, PYTHIA-II, that makes software recommendations based on problem specifications and computational objectives such as accuracy, cost or time, and memory requirements.It is designed to (i) identify and select the software/hardware resources available for a user's problem, (ii) locate these resources and provide information about their usage, availability, cost and related information, (iii) suggest parameter values, and (iv) provide an assessment of the recommendation.In addition, PYTHIA-II can be used to generate "testing" software repositories, since it provides all the necessary facilities to set up database schemas for testing benchmarks and associated performance data, with a number of tools for visualization, statistical ranking, data mining, knowledge representation, and recommendation generation.