(Swansea)

Approximations based on dyadic centred intervals are investigated as a meansfor implementing exact real arithmetic. It is shown that the field operations

can be implemented on these approximations with optimal or near optimal

results. Bounds for the loss in quality of approximations for each of the

field operations are also given. These approximations can be used as a more

efficient alternative to endpoint based implementations of interval analysis.

Board Room - - in cooperation with the PCV Seminar

Department of Computer Science