In recent years, lots of multifactorial optimization evolutionary algorithms have been developed to optimize multiple tasks simultaneously, which improves the overall efficiency using implicit genetic complementarity between different tasks. In this paper, a novel multitask fireworks algorithm is proposed with novel transfer sparks to solve multitask optimization problems. For each task, some transfer sparks would be generated with adaptive length and promising direction vector, which are very helpful to transfer useful genetic information between different tasks. Finally, the proposed algorithm is compared against some chosen state-of-the-art evolutionary multitasking algorithms. The experimental results show that the proposed algorithm provides better performance on several single objectives and multiobjective MTO test suites.

中文翻译：

---

基于转移火花的Fireworks进化多任务处理算法。

近年来，已经开发了许多多因素优化进化算法来同时优化多个任务，这利用不同任务之间的隐式遗传互补性提高了整体效率。为了解决多任务优化问题，提出了一种新颖的带有多任务传递火花的多任务烟花算法。对于每个任务，将生成具有自适应长度和有希望的方向向量的转移火花，这对于在不同任务之间转移有用的遗传信息非常有帮助。最后，将所提出的算法与一些选定的最新发展型多任务处理算法进行比较。实验结果表明，该算法在多个单目标和多目标MTO测试套件上具有更好的性能。