Numerical program analysis and testing

Abstract

Numerical software is playing an increasingly critical role in modern society, but composing correct numerical programs is difficult. This paper describes a doctoral research program that aims to alleviate this issue. It tackles real world problems and is guided by features learned from empirically studying these programs. By assisting developers in the production of numerical software, it improves the quality and productivity of software development. The research depends on numerical analysis and lies in the intersection of software engineering and program analysis.