So ware veri cation competitions, such as the annual SV-COMP, evaluate so ware veri cation tools with respect to their e ectivity and e ciency. Typically, the outcome of a competition is a (possibly category-speci c) ranking of the tools. For many applications, such as building portfolio solvers, it would be desirable to have an idea of the (relative) performance of veri cation tools on a given veri cation task beforehand, i.e., prior to actually running all tools on the task.In this paper, we present a machine learning approach to predicting rankings of tools on veri cation tasks. e method builds upon so-called label ranking algorithms, which we complement with appropriate kernels providing a similarity measure for veri cation tasks. Our kernels employ a graph representation for so ware source code that mixes elements of control ow and program dependence graphs with abstract syntax trees. Using data sets from SV-COMP, we demonstrate our rank prediction technique to generalize well and achieve a rather high predictive accuracy. In particular, our method outperforms a recently proposed feature-based approach of Demyanova et al. (when applied to rank predictions).
CCS CONCEPTS•Computing methodologies →Ranking; Support vector machines; Cross-validation; •So ware and its engineering →So ware veri cation; Formal so ware veri cation;
KEYWORDSSo ware veri cation, machine learning, ranking.