An underpinning numerical software environment for metrology?
Ian Reid
The Numerical Algorithms Group (NAG) Limited\\
Wilkinson House, Jordan Hill Road, OXFORD\\
Friday, August 15, 2003 11:00-12:00,
Lecture Room C in Building 101 (NIST Admin)
Gaithersburg
Friday, August 15, 2003 09:00-10:00,
Room 4550
Boulder
Abstract:
Metrologists should not have to be concerned about the quality and accuracy
of their underlying numerical software. Of course, as with any calculation,
one should have an estimate of the answer, but there should be no need to
check that the software is working to the accuracy claimed. However, some
excellent work at NPL in "Testing Spreadsheet and Other Packages Used
in
Metrology" as part of the Software Support for Metrology (SSfM) programme,
demonstrated that some of the tools in common use within Metrology and other
disciplines are lacking in several areas with regard to accuracy. This, of
course, is and should be of great concern to those using these tools and
relying on their results.
One approach to this problem would be to have a central body work to
validate each environment used by Metrologists and to warn of areas
(environments, or specific parts of environments) to be avoided. Given
limited resources, it would make sense to reduce the number of environments
which are validated. Some extremely valuable work in this area is on-going
under the Validation and Testing theme of the SSfM-2 programme.
The METROlogy Software environment
might be seen as an embodiment of this
approach. METROS is a web-based delivery platform designed to give
metrologists access to the numerical software they need to solve their
problems, with a level of validation associated with each solution. This
talk will expose the philosophy behind METROS and the progress towards its
implementation.
Another approach would be to validate a set of components which could be
called from a range of common environments. This approach would
significantly reduce the amount of validation required, whilst allowing the
Metrologists to remain within their familiar/favoured environments. This
approach is being championed by the Centre for Mathematics and Scientific
Computing (CMSC) at NPL. The components chosen for this approach are those
available as the NAG Libraries. This talk will look briefly at some of the
problems uncovered within the SSfM programme and look at the reasons why
NPL, along with a whole range of major research communities, commercial
clients and independent software vendors have turned to NAG to provide
numerical and statistical software components.
Contact: A. J. KearsleyNote: Visitors from outside NIST must contact
Robin Bickel; (301) 975-3668;
at least 24 hours in advance.
|
