The concept of channel, a computational mechanism used to convey state to different threads of process execution, is at the core of the design of multi-threaded concurrentalgorithms. In the case of concurrent evoluti...
详细信息
ISBN:
(纸本)9783030310196;9783030310189
The concept of channel, a computational mechanism used to convey state to different threads of process execution, is at the core of the design of multi-threaded concurrentalgorithms. In the case of concurrent evolutionary algorithms, channels can be used to communicate messages between several threads performing different evolution tasks related to genetic operations or mixing of populations. In this paper we study to what extent the design of these messages in a communicating sequential process context may influence scaling and performance of concurrent evolutionary algorithms. For this aim, we designed a channel-based concurrentevolutionary algorithm that is able to effectively solve different benchmark binary problems (e.g. OneMax, LeadingOnes, RoyalRoad), showing that it provides a good basis to leverage the multithreaded and multi-core capabilities of modern computers. Although our results indicate that concurrency is advantageous to scale-up the performance of evolutionaryalgorithms, they also highlight how the trade-off between concurrency, communication and evolutionary parameters affect the outcome of the evolved solutions, opening-up new opportunities for algorithm design.
concurrent evolutionary algorithms use threads that communicate via messages. Parametrizing the work in every thread and the way they communicate results is a major challenge in its design. In this paper we work with ...
详细信息
ISBN:
(纸本)9781450367486
concurrent evolutionary algorithms use threads that communicate via messages. Parametrizing the work in every thread and the way they communicate results is a major challenge in its design. In this paper we work with concurrent evolutionary algorithms implemented in Perl 6, and explore different options of single-thread evolution parametrization, communication and mixing of results, showing that scalability is achieved in a multi-core environment.
暂无评论