| Research |
PELLPACK
is a PSE which supports the process of
modeling physical applications described by
Partial Differential equations (PDEs).
PELLPACK consists of an interactive graphical interface which supports
symbolic processing, equation editors, PDE model specification,
geometry tools, analysis tools, mesh generators,
geometry partioning tools, visualization packages, PDE solving libraries,
integrated foreign systems, machine configuration, and parallel processing.
A web-based demonstration has been established at
WebPDELab
For an overview of the web project, click
here.
The PELLPACK network server provides a powerful internet based problem
solving facility. Users anywhere on the internet can access it to
solve a broad range of simulations for physical objects and processes.
PYTHIA-II applies KDD to the study of scientific software for the
solution of PDEs. The result of PYTHIA-II research is a
Recommender system which answers user queries regarding the
applicability, efficiency and accuracy of software for solving
PDE problems. Supported by an inference process, the
PYTHIA-II Recommender assimilates information from
knowledge bases built
by the data mining of domain specific performance data.
Once the computational objectives for the performance evaluation
of the selected scientific software have been determined,
the KDD process is used to generate a knowledge base as
shown in the table below. The implementation of PYTHIA-II includes
a complete graphical user interface supported by the
POSTGRES95 RDBMS, in-house
statistical software,
the GOLEM ILP learning system and
the CLIPS expert system tool-box.
| Data Preparation |
select benchmark applications (problem population) |
identify performance measures | select methods (libraries/routines & parameters) |
use domain-specific schema to populate database |
| Data Generation |
use database population to build PDE 'programs' |
execute programs, generating performance output | collect and store performance data | |
| Data Mining |
prepare/extract/filter/clean data | apply statistical techniques | generate profiles to characterize methods and parameters | identify patterns/models from analyzed data |
Look at the preliminary paper PYTHIA-II: A KDD Based Recommender System for Scientific Computing (8MB postscript file) for more information about PYTHIA-II.
PDELab
is a problem solving and development environment for PDE based
applications. PDELAB
introduces the PDESpec symbolic language which invokes
Mathematica for symbolic processing.
The PDELAB Notebook, Object Manager, Editor Toolkits
and PDE solver components use
PDE object representations as their foundation,
and the PDELAB environment is implemented using the
Purdue PSE Kernel (PPK).
The PDELAB components can be wired together at a high level
to form a custom
PDE solver for a given physical application. PDELAB allows users to
introduce new PDE solver modules and build high level templates to study
the behavior of solution processes.
SoftLab is a framework for building virtual laboratories for
computational science. A virtual laboratory is a hybrid PSE which
encompasses all research activities that occur in the scientific
laboratory environment. It is a software layer
above the experimental and computational processes which generally
exist in the laboratory side-by-side. The virtual
laboratory unites the experimental and computational models
in a way that allows them to interact with each other,
so that feedback from one can
enhance and improve the methodologies applied in the other.
Visit the Problem Solving Environments Research homepage.
(*) If your web server cannot handle a postscript file, use shift+button1 to download the file. Then use the software facilities at your site to print or view it with a postscript viewer.