论文部分内容阅读
随着工程应用越来越复杂,待优化参数指数增长,原有的参数优化算法丧失了有效性,基于此提出一种基于图形处理器的并行小波突变协同差分进化算法.该方法首先依据协同进化策略构造协同差分进化框架,将高维问题分解为一些子种群,并同时将总问题分配到每个图形处理单元中,进而在子种群内利用小波突变差分进化算法进行同步并行计算,提高全局持续搜索能力.通过高维标准函数对比测试分析,所提算法相对其他方法具有较好的参数优化能力和计算效率.
With the complexity of engineering application and the exponential increase of parameters to be optimized, the original parameter optimization algorithm has lost its validity. Based on this, a parallel differential evolution mutation algorithm based on GPU is proposed. The strategy constructs a cooperative differential evolution framework, which decomposes the high-dimensional problem into some sub-populations and allocates the total problem to each GPU simultaneously, and then uses the wavelet mutation differential evolution algorithm to carry out synchronous and parallel computation in the sub-population to improve the global persistence Search ability.According to high-dimensional standard function comparison test analysis, the proposed algorithm has better parameter optimization ability and computational efficiency than other methods.