排队"梯度"浮点数编码遗传算法

文章主要论述了浮点数编码遗传算法优化的问题。充分利用个体及其适应度函数值的已知信息,依据一定算法得到“梯度”信息,从而加快算法的收敛速度。但如果个别适应度函数值相对较大(例如是全体适应度函数值平均值的100倍以上),则上述利用“梯度”信息的优势尽失。为了解决此问题,提出排队“梯度”浮点数编码遗传算法。最后用典型的四个常用测试函数分别对简单浮点数编码遗传算法、利用“梯度”信息的浮点数编码遗传算法和排队“梯度”浮点数编码遗传算法的性能进行了测试和比较。

Queued "Gradient" Float-encoded Genetic Algorithm

This paper mainly discusses about the optimization of the float-encoded genetic algorithm. We can make the best of known info of the individuals and their fitness function's value. Then we can get the so-called "gradient" about them according to some algorithm to speed up the constringency rate of the algorithm. But if some fitness function value is relatively bigger (for example it is 100 times than the average value) the merit using the "gradient" info may be lost drastically. To solve the problem we bring forward queued "gradient" float-encoded genetic algorithm. At last we use four normal testing functions to test and compare the performance of the simple float-encoded genetic algorithm, using "gradient" info float-encoded genetic algorithm and queued "gradient" float-encoded genetic algorithm respectively.