进化计算的二进制与浮点数自适应混合编码方法

Evolutionary Computation with Adaptive Hybrid Binary and Real Encoding Method

提出了一种新的基于二进制码串和浮点数码串的混合编码方法，将遗传个体的每一变量编码为二进制和浮点数的混合码串，使得进化计算能结合二进制编码方法全局探索能力强的特点和浮点数编码方法局部寻优能力强的特点。在这种混合编码方法中，根据进化过程中遗传群体平均适应度的变化，自适应地改变二进制码串和浮点数码串所占比例，使得进化计算的全局探索能力和局部寻优能力得到了较好的平衡，提高了进化计算的收敛速度和全局优化收敛率。文章对一系列典型函数的优化计算实验验证了基于二进制码串和浮点数码串的混合编码的特点。

A novel hybrid encoding method based on binary and real codes is proposed in this paper. As well known, the evolutionary computation (EC) with binary codes presents stronger global exploration ability and the EC with real codes has stronger local search ability. The proposed method combines the two advantages by encoding each variable of genetic individual into a hybrid binary-real string. The ratio of binary code and real code in one string is adaptively changed according to the average fitness change of genetic population during the evolutionary procedure, so the proposed method gains the good trade-off between the global exploration and local exploitation, and obtains fast convergence speed and global convergence rate. The experiments on optimizing of a series of typical test functions also verified the good performance.