Concurrently searching branches in software tests generation through multitask evolution

Multitask evolutionary computation (MT-EC) has been recently identified as a potentially useful paradigm for significant real-world domains. One such domain is the field of software testing. Although a number of evolutionary approaches exist already, there is a lack of strategies that can leverage knowledge from different sources to enhance the search process. In this work, we focus on branch testing and explore the capability of MT-EC to guide the search by exploiting inter-branch information. To the best of our knowledge, this is the first application of MT-EC to real-world problems with more than two tasks. Precisely, we evince that selection, together with the preference relation used to compare individuals, form a mechanism capable of achieving a concurrent search for the branches while exploiting inter-branch knowledge in the process. Further, we demonstrate that the intensity of the transfer can be altered with the implemented selection. The experimental results on benchmark programs suggest that MT-EC can be specially useful in situations where the budget for the search process is limited.