This paper presents an improvement to ECJ (Evolutionary Computation in Java), the popular evolutionary computation tool, which allows users to exploit distributed computational resources through the use of volunteer computing. In particular, the BOINC (Berkeley Open Infrastructure for Network Computing) middleware is used to distribute ECJ client software on top of a virtualization layer, this allows researchers to parallelize their experiments without having to port their software to make it compatible with BOINC. In this way, an interested researcher can use the ECJ+BOINC software in the same way that they use the standard ECJ tool. Moreover, the system allows the user to choose be-tween different distribution models based on their preferences. Finally, the ECJ+BOINC system is developed in a modular manner, allowing for easy updates and modifications based on the possible requirements of future ECJ versions.
- A Hybrid ECJ+BOINC Tool for Distributed Evolutionary Algorithms.pdf