The Error-Analyst's Quandary.
Abstract
Numerical analysts often pass the buck by alleging that certain computational schemes are 'numerically stable' even when they produce palpably wrong answers. The note is intended first to help the layman understand why those allegations, however misleading, may be true, and second to show numerical analysts that the buck is not so easy to pass as might at first appear. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1972
- Accession Number
- AD0753856
Entities
People
- W. Kahan
Organizations
- University of California, Berkeley