GEP算法解码结构复用研究

Study on reutilization of GEP algorithm decoding structure

适应度评价大体可以分成解码和计算适应度值两个部分,是进化计算中运算量最大、重复率最高的过程之一。为了有效利用已有计算结果的角度避免大量重复建树和遍历运算,改进了GEP解码算法基本流程以降低GEP的运算量,达到了提升运算效率的目的。采用仿真的方式对引入复用机制的GEP和传统GEP算法进行了对比实验,从解码次数、解码所用时间等两个方面进行对照,发现引入复用机制的GEP算法在解码次数上比传统GEP有多个数量级的优势,在解码所需时间上也有较大的改进。

Fitness evaluation can be generally divided into two parts：decoding and calculating the fitness value. It concludes the largest operation amount and the highest repetition rate in evolutionary computation. From the perspective of the effective utilization of the acquired calculation results to avoid repetitive builds and traversal operation, this paper achieved the goal of reducing the GEP computation and enhancing operation efficiency by improving the basic flow of GEP decoding algorithm. At last, according to simulation method, it conducted a comparative experiment between the GEP algorithm which introduced the reuse mechanism and the original one. The comparison mainly focused on respective decoding times and time. It discovered that via introducing the reuse mechanism into GEP algorithm, the algorithm could significantly reduce the decoding times by several orders of magnitude and save the decoding time to some extent than the original one.