On a theoretical justification of the choice of epsilon-inflation in PASCAL-XSC (Q1375604)

The authors show that the so-called epsilon-inflation of an interval \([x^-,x^+]\) given by the formula \[ [x^-,x^+]_{\varepsilon}=[(1+\varepsilon)x^--\varepsilon x^+,(1+\varepsilon)x^+-\varepsilon x^-], \] which was implemented in PASCAL-XSC, is invariant with respect to some transformations and is in certain sense optimal.