Summary: This minisymposium is concerned with a central problem in scientific computing, that of assessing the accuracy of solutions obtained by numerical methods on computers using floating point arithmetic. The speakers in this minisymposium will review different methods and approaches to solving this problem.
The first speaker will look at an approach called Qualitative Computing, the second speaker will look at the state of the art in rounding error analysis, the third speaker will offer a critique of the probabilistic approach to estimating roundoff error, and the fourth speaker will review the methods of interval analysis.
Day: July 5, 1995 Time: 09.30
F. Chaitin-Chatelin, Université Paris IX Dauphine and CERFAS, France:
Qualitative computing and the robustness of numerical methods to
high nonnormality
J. Demmel, University of California at Berkeley, USA:
Modern error analysis in the LAPACK linear algebra library
W. Kahan, University of California, USA:
The improbability of probabilistic estimates for roundoff
G. Mayer, Universität Rostock, Germany:
Interval methods for the assessment of numerical accuracy