一种新的基于量子遗传算法的ECOC算法  

A New ECOC Algorithm Based on the Quantum Inspired Genetic Algorithm

作　　者：周大鹏 ZHOU Dapeng(School of Computing and Artificial Intelligence,Southwest Jiaotong University,Chengdu 611756,China)

机构地区：[1]西南交通大学计算机与人工智能学院,四川成都611756

出　　处：《现代信息科技》2023年第10期22-25,共4页Modern Information Technology

摘　　要：纠错输出编码(ECOC)将多分类问题转化为二类问题进行求解。其中,影响ECOC性能的关键因素是最优编码矩阵,为构建有效的最优编码矩阵,文章提出一种新的基于量子遗传算法的ECOC算法。首先,将ECOC矩阵作为量子遗传算法中的个体,使用量子位编码重新生成编码矩阵。随后,利用交叉、变异、量子旋转门等遗传算子,使ECOC算法朝着最优的方向进化。在12个标准UCI数据集上进行的实验表明所提出算法具有良好的分类性能。Error Correcting Output Codes(ECOC)transforms the multi-class classification problems into the two-class problems to solve.The key factor affecting the performance of the ECOC is the optimal coding matrix.In order to construct an efficient and optimal coding matrix,a new ECOC algorithm based on Quantum Inspired Genetic Algorithm is proposed in this paper.Firstly,one ECOC coding matrix is regarded as one individual in Quantum Inspired Genetic Algorithm,and the coding matrix is reconstructed by the q-bit coding.Then,the genetic operators such as crossover,mutation,quantum rotating gate are used to make the ECOC algorithm evolve toward the optimal direction.Experiments conducted on 12 standard UCI data sets show that the proposed algorithm has better classification performance.

关 键 词：多分类 纠错输出编码 量子遗传算法 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]