CAJ | 학술논문

为提高MOSES效率,提出了一种新的程序树层次化结构统计模型.该模型通过统计分析同类群,自动发现子树特征来指导优化.该模型不需要hBOA算法那样对变量集合进行建模,也不需要像MRTS算法那样遍历小规模的种群来发现潜在的有指导意义的子树.通过解决人工蚂蚁问题对算法进行了测试,结果表明改进后的MOSES算法更加高效.
위제고MOSES효솔,제출료일충신적정서수층차화결구통계모형.해모형통과통계분석동류군,자동발현자수특정래지도우화.해모형불수요hBOA산법나양대변량집합진행건모,야불수요상MRTS산법나양편력소규모적충군래발현잠재적유지도의의적자수.통과해결인공마의문제대산법진행료측시,결과표명개진후적MOSES산법경가고효.
To improve the efficiency of MOSES algorithm, this paper proposes a new hierarchical statistical model of program trees. This model conducts hierarchical statistical analysis on program trees and can generate potential subtrees automatically to guide algorithm optimization. This model leaves out the operations of creating models for the variables set like the previous hBOA algorithm; and also doesn't need the tedious operations to traversal small population to find certain superior individuals as subtrees like the MRTS method. Experimental results on solving artificial ant problem indicate that our proposed algorithm is more effective and efficient than the previous hBOA-based MOSES.