中南大学学报(自然科学版)
中南大學學報(自然科學版)
중남대학학보(자연과학판)
JOURNAL OF CENTRAL SOUTH UNIVERSITY
2009年
5期
1345-1353
,共9页
人工免疫%优化%免疫网络算法%禁忌搜索算法
人工免疫%優化%免疫網絡算法%禁忌搜索算法
인공면역%우화%면역망락산법%금기수색산법
artificial immune%optimization%artificial immune network algorithm%tabu search algorithm
基于人工免疫网络算法(aiNet)模型,借鉴禁忌搜索算法(TS)的思想,提出一种禁忌搜索与人工免疫的混合算法,即人工免疫网络算法(TS-aiNet).在算法中引入禁忌表,禁忌那些在网络迭代中亲和度连续不再增加的细胞,并通过特赦准则赦免一些被禁忌的优良状态;增加1个记忆表,用于保存成熟的记忆细胞;重新定义高斯变异方式,以保证多样化的有效搜索.利用Markov链证明算法全局收敛性,通过对多个典型系统测试函数的仿真实验定量分析该算法的性能,并与经典克隆选择算法和opt-aiNet算法进行比较研究,分析特征参数对算法性能的影响.实验结果表明,该算法在多模态搜索空间中具有更强的全局收敛性、稳定性和寻找极值点能力,能够克服早熟现象,是一种有效的全局优化搜索方法.
基于人工免疫網絡算法(aiNet)模型,藉鑒禁忌搜索算法(TS)的思想,提齣一種禁忌搜索與人工免疫的混閤算法,即人工免疫網絡算法(TS-aiNet).在算法中引入禁忌錶,禁忌那些在網絡迭代中親和度連續不再增加的細胞,併通過特赦準則赦免一些被禁忌的優良狀態;增加1箇記憶錶,用于保存成熟的記憶細胞;重新定義高斯變異方式,以保證多樣化的有效搜索.利用Markov鏈證明算法全跼收斂性,通過對多箇典型繫統測試函數的倣真實驗定量分析該算法的性能,併與經典剋隆選擇算法和opt-aiNet算法進行比較研究,分析特徵參數對算法性能的影響.實驗結果錶明,該算法在多模態搜索空間中具有更彊的全跼收斂性、穩定性和尋找極值點能力,能夠剋服早熟現象,是一種有效的全跼優化搜索方法.
기우인공면역망락산법(aiNet)모형,차감금기수색산법(TS)적사상,제출일충금기수색여인공면역적혼합산법,즉인공면역망락산법(TS-aiNet).재산법중인입금기표,금기나사재망락질대중친화도련속불재증가적세포,병통과특사준칙사면일사피금기적우량상태;증가1개기억표,용우보존성숙적기억세포;중신정의고사변이방식,이보증다양화적유효수색.이용Markov련증명산법전국수렴성,통과대다개전형계통측시함수적방진실험정량분석해산법적성능,병여경전극륭선택산법화opt-aiNet산법진행비교연구,분석특정삼수대산법성능적영향.실험결과표명,해산법재다모태수색공간중구유경강적전국수렴성、은정성화심조겁치점능력,능구극복조숙현상,시일충유효적전국우화수색방법.
A hybrid approach, tabu search artificial immune algorithm (TS-aiNet) was proposed based on aiNet model inspired by mechanism of tabu search algorithm. A tabu list was introduced to taboo such cell whose affinities didn't continuously increase any more in the network. In some phrases the tabooed excellent cells were released according to aspiration criteria. To save mature memory cells, a memory table was added to cells network. In addition, the expression of Gauss mutation was redefined for diversity search in the process of global optimization. Markov chain was applied to prove global convergence. Performance analysis of optimization was carried out based on random simulation of some typical systems, which was compared with that of KLONALG and opt-aiNet algorithms. Finally, the influence of feature parameters on TS-aiNet algorithm was analyzed. The simulation results show that the presented approach has preferable global convergent ability and stability in multi-modal search space, and can avoid prematurity effectively. So it is demonstrated as a global optimized algorithm with feasibility and high efficiency.