Attract-Repulse Fireworks Algorithm and its CUDA Implementation Using Dynamic Parallelism

• DOI: 10.4018/ijsir.2015040101
• Publication Date: 2015-07-16T14:08:26Z
• Authors: Ke Ding; Ying Tan
• Keywords: 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

Fireworks Algorithm (FWA) is a recently developed Swarm Intelligence (SIA), which has been successfully used in diverse domains. When applied to complicated problems, many function evaluations are needed obtain an acceptable solution. To address this critical issue, GPU-based variant (GPU-FWA) was proposed greatly accelerate the optimization procedure of FWA. Thanks active studies on FWA and GPU computing, advances have achieved since GPU-FWA. In paper, novel variant, Attract-Repulse (AR-FWA), proposed. AR-FWA introduces efficient adaptive search mechanism (AFW Search) non-uniform mutation strategy for spark generation. Compared state-of-the-art variants, can improve performance multimodal problems. Leveraging edge-cutting dynamic parallelism provided by CUDA, be implemented easily efficiently.