Termination Analysis by Learning Terminating Programs

Abstract

We present a novel approach to termination analysis. In a first step, the analysis uses a program as a black-box which exhibits only a finite set of sample traces. Each sample trace is infinite but can be represented by a finite lasso. The analysis can "learn" a program from a termination proof for the lasso, a program that is terminating by construction. In a second step, the analysis checks that the set of sample traces is representative in a sense that we can make formal. An experimental evaluation indicates that the approach is a potentially useful addition to the portfolio of existing approaches to termination analysis.