第 40 卷 第 9 期 自 动 化 学 报 Vol. 40, No. 9
2014 年 9 月 ACTA AUTOMATICA SINICA September, 2014
基于反馈的精英教学优化算法
于坤杰
1
王 昕
2
王振雷
1
摘 要 精英教学优化算法 (Elitist teaching-learning-based optimization, ETLBO) 是一种基于实际班级教学过程的新型
优化算法. 本文针对 ETLBO 算法寻优精度低、稳定性差的问题, 提出了反馈精英教学优化算法 (Feedback ETLBO). 在
ETLBO 算法的基础上, 通过在学生阶段之后加入反馈阶段, 增加了学生的学习方式, 保持学生的多样性特性, 提高算法的全
局搜索能力. 同时, 反馈阶段是选举成绩较差的学生与教师交流, 使成绩较差的学生快速向教师靠拢, 使算法进行局部精细搜
索, 提高算法的寻优精度. 对 6 个无约束及 5 个约束标准函数的测试结果表明, FETLBO 算法与其他算法相比在寻优精度和
稳定性上更具优势. 最后将 FETLBO 算法应用于拉压弹簧优化设计问题及 0-1 背包问题, 取得了满意结果.
关键词 进化算法, 精英教学优化算法, 反馈, 函数优化
引用格式 于坤杰, 王昕, 王振雷. 基于反馈的精英教学优化算法. 自动化学报, 2014, 40(9): 1976−1983
DOI 10.3724/SP.J.1004.2014.01976
Elitist Teaching-learning-based Optimization Algorithm Based on Feedback
YU Kun-Jie
1
WANG Xin
2
WANG Zhen-Lei
1
Abstract Elitist teaching-learning-based optimization (ETLBO) is a novel optimization algorithm based on the practical
teaching-learning process of the class. In this paper, we propose a feedback elitist teaching-learning-based optimization
(FETLBO) to solve the problem of low precision and poor stability of the ETLBO. Based on the ETLBO, a feedback
phase is introduced at the end of the learner phase to increase the learning style and ensure the diversity of students so as
to improve the algorithm
0
s global search ability. Meanwhile, the feedback phase is for the slow students to communicate
with the teacher and enables them to be close to the teacher quickly, so that the algorithm uses the fine local search and
improves the precision. Six unconstrained and five constrained classic tests show that the FETLBO algorithm outperforms
the other algorithms in precision and stability. Finally, the FETLBO algorithm is applied to the tension/compression
spring design problem and the 0-1 knapsack problem, and obtains satisfactory results.
Key words Evolutionary algorithms, elitist teaching-learning-based optimization algorithm, feedback, function opti-
mization
Citation Yu Kun-Jie, Wang Xin, Wang Zhen-Lei. Elitist teaching-learning-based optimization algorithm based on
feedback. Acta Automatica Sinica, 2014, 40(9): 1976−1983
收稿日期 2013-07-08 录用日期 2014-02-26
Manuscript received July 8, 2013; accepted February 26, 2014
国家重点基础研究发展计划 (973 计划) (2012CB720500), 国家自
然科学基金 (61333010, 21276078, 21206037), 中央高校基本科研
业 务 费 专 项 资 金 (863 计 划) (2013AA0400701), 上 海 市 科 技 攻 关
(12dz1125100), 十二五国家科技支撑计划 (2012BAF05B00), 上海
市重点学科建设项目 (B504), 上海市自然科学基金 (14ZR1421800),
流程工业综合自动化国家重点实验室开放课题基金资助项目 (PAL-
N201404) 资助
Supported by National Basic Research Program of China (973
Program) (2012CB720500), National Natural Science Founda-
tion of China (61333010, 21276078, 21206037), The Central
University Basic Scientific Research Business Expenses Spe-
cial Funds (863 Program) (2013AA0400701), Shanghai Science
and Technology Research Projects (12dz1125100), National Sci-
ence and Technology Support Project during the 12th Five-Year
Plan Period (2012BAF05B00), Shanghai Leading Academic Dis-
cipline Project (B504), Shanghai Natural Science Foundation
(14ZR1421800), the State Key Laboratory of Synthetical Au-
tomation for Process Industries (PAL-N201404)
本文责任编委 刘德荣
Recommended by Associate Editor LIU De-Rong
1. 华东理工大学化工过程先进控制和优化技术教育部重点实验室 上海
200237 2. 上海交通大学电工与电子技术中心 上海 200240
1. Key Laboratory of Advanced Control and Optimization for
在工程优化、计算机网络和人工智能等领域
常常会遇到大规模、非线性、多极值等问题, 这类
问题很难用传统数学手段处理. 智能优化算法依
据计算机的迭代计算能力, 不依赖优化问题本身,
很好地解决了这一问题. 从 1975 年遗传算法 (Ge-
netic algorithm, GA)
[1]
提出至今, 涌现出许多智
能优 化算法, 如粒 子群算法 (Particle swarm op-
timization, PSO)
[2−5]
、差分进化算法 (Differential
evolution, DE)
[6−7]
、群搜索算法 (Group search op-
timization, GSO)
[8]
和人工蜂群算法 (Artificial bee
colony, ABC)
[9−11]
等. 然而这些算法在搜索和寻优
性能方面还存在一些缺陷, 例如粒子群算法的求解
是多样性逐渐丧失的过程, 所以粒子群算法局部寻
优能力较差且易于发生早熟收敛现象. 因此, 研究者
们将一些新理论、新方法引入其中, 为智能算法的发
Chemical Processes, East China University of Science and Tech-
nology, Shanghai 200237 2. Center of Electrical and Electronic
Technology, Shanghai Jiao Tong University, Shanghai 200240