本文選題：貝葉斯 + L型TSK模糊系統(tǒng)��； 參考：《控制與決策》2017年10期

【摘要】：針對傳統(tǒng)Takagi-Sugeno-Kan(TSK)模糊系統(tǒng)處理大規(guī)模數(shù)據(jù)時間代價較高的問題,提出一種基于概率模型框架的L2型TSK模糊系統(tǒng)建模策略,建立具有處理大規(guī)模數(shù)據(jù)能力的貝葉斯L2型TSK模糊系統(tǒng)(B-TSK-FS).具體地,基于L2型TSK模糊系統(tǒng)的輸出誤差概率化表示,對系統(tǒng)前后件參數(shù)聯(lián)合學習,提高系統(tǒng)的泛化能力.另外,引入狄利克雷先驗分布函數(shù),對模糊隸屬度稀疏化表示,實現(xiàn)樣本的壓縮,降低運算時間.在模擬和真實數(shù)據(jù)集上的實驗結果驗證了所提出模糊系統(tǒng)的優(yōu)勢.
[Abstract]:In order to solve the problem of high time cost for traditional Takagi-Sugeno-Kanko fuzzy system to deal with large-scale data, a modeling strategy of L2 type TSK fuzzy system based on probabilistic model framework is proposed, and a Bayesian L2 type TSK fuzzy system with large scale data processing capability is established. Specifically, based on the probabilistic representation of the output error of L2 type TSK fuzzy system, the parameters of the front and rear parts of the system are jointly studied to improve the generalization ability of the system. In addition, the prior distribution function of Delikley is introduced to sparse representation of fuzzy membership degree, which can compress the sample and reduce the operation time. Experimental results on simulated and real data sets verify the advantages of the proposed fuzzy system.
【作者單位】： 江南大學數(shù)字媒體學院;中國科學院深圳先進技術研究院廣東省機器視覺與虛擬現(xiàn)實技術重點實驗室;
【基金】：國家自然科學基金項目(61300151,61305097) 深圳市基礎學科布局項目(JCYJ20160429190300857)
【分類號】：O159;O212.8
，

本文編號：1791978