Implementation of affine arithmetic using floating-point arithmetic: how to handle roundoff errors
Résumé
Affine arithmetic is one of the extensions of interval arithmetic, which aims
at counteracting the variable dependency problem. The focus of our work is
the implementation of affine arithmetic using floating-point arithmetic. More
precisely, roundoff errors are properly accounted for in our implementation. All
coefficients of affine forms are floating-point numbers. A noise symbol ν is
devoted to the handling of roundoff errors.