计算机应用与软件
計算機應用與軟件
계산궤응용여연건
Computer Applications and Software
2015年
8期
308-314
,共7页
分布估计算法%模糊规则优化%分布概率模型%混杂优化算法
分佈估計算法%模糊規則優化%分佈概率模型%混雜優化算法
분포고계산법%모호규칙우화%분포개솔모형%혼잡우화산법
Distribution estimation algorithm%Fuzzy rules optimisation%Distribution probability model%Hybrid optimisation algorithm
为了降低当前模糊规则优化学习算法的时间复杂度，加快其收敛速度，基于单变量边缘分布估计算法，引入CH（ Cordon and Herrera）与COR（ Cooperative Rules Methodology）机制，提出混杂分布估计算法耦合CH（ Cordon and Herrera）与COR联合机制的模糊规则优化算法研究；对算法的时间复杂度进行理论推导和分析证明，构建算法的分布概率模型。首先使用CH机制产生变量空间；再由COR方法完备的候选规则库；然后利用多种群变量无关分布估计算法MUMDA（ Univariate Marginal Distribution Algorithm）进行规则学习，通过增加种群的多样性，减少算法陷入局部最优解的可能；最后对该算法进行实验验证。实验对比结果可以看出，该设计的混杂优化算法的计算，可获得精度较高、可理解性较强的模糊规则库，便于模糊系统在实际工程中的应用。
為瞭降低噹前模糊規則優化學習算法的時間複雜度，加快其收斂速度，基于單變量邊緣分佈估計算法，引入CH（ Cordon and Herrera）與COR（ Cooperative Rules Methodology）機製，提齣混雜分佈估計算法耦閤CH（ Cordon and Herrera）與COR聯閤機製的模糊規則優化算法研究；對算法的時間複雜度進行理論推導和分析證明，構建算法的分佈概率模型。首先使用CH機製產生變量空間；再由COR方法完備的候選規則庫；然後利用多種群變量無關分佈估計算法MUMDA（ Univariate Marginal Distribution Algorithm）進行規則學習，通過增加種群的多樣性，減少算法陷入跼部最優解的可能；最後對該算法進行實驗驗證。實驗對比結果可以看齣，該設計的混雜優化算法的計算，可穫得精度較高、可理解性較彊的模糊規則庫，便于模糊繫統在實際工程中的應用。
위료강저당전모호규칙우화학습산법적시간복잡도，가쾌기수렴속도，기우단변량변연분포고계산법，인입CH（ Cordon and Herrera）여COR（ Cooperative Rules Methodology）궤제，제출혼잡분포고계산법우합CH（ Cordon and Herrera）여COR연합궤제적모호규칙우화산법연구；대산법적시간복잡도진행이론추도화분석증명，구건산법적분포개솔모형。수선사용CH궤제산생변량공간；재유COR방법완비적후선규칙고；연후이용다충군변량무관분포고계산법MUMDA（ Univariate Marginal Distribution Algorithm）진행규칙학습，통과증가충군적다양성，감소산법함입국부최우해적가능；최후대해산법진행실험험증。실험대비결과가이간출，해설계적혼잡우화산법적계산，가획득정도교고、가리해성교강적모호규칙고，편우모호계통재실제공정중적응용。
In order to reduce the time complexity of current fuzzy rules optimisation learning algorithm and to speed up the convergence rate, based on univariate marginal distribution estimation algorithm, we introduced CH ( Cordon & Herrera) and COR ( cooperative rules) mechanisms, and presented the study of fuzzy rules optimisation algorithm which couples the hybrid distribution estimation algorithm with the CH and COR joint mechanism;Moreover we carried out the theoretical derivation and analytical demonstration on the time complexity of the algorithm, and built the distribution probability model of the algorithm.First, we employed CH mechanism to generate variable space, and completed the candidate rules library with COR method.Then, we used the MUMDA ( multi-population variable irrelative distribution estima-tion algorithm) for rules learning, by increasing the diversity of population the possibility of the algorithm falling into local optimal was dimin-ished.Finally we conducted the experimental validation on the algorithm, it is shown by the experimental comparison result that the hybrid optimisation algorithm designed in the paper could obtain a fuzzy rules library with high accuracy and better comprehensibility, and this facili-tated the fuzzy system to be applied in practical projects.
