具備視角協(xié)同學(xué)習(xí)能力的多視角TSK型模糊系統(tǒng)
doi: 10.11999/JEIT151209
基金項(xiàng)目:
國(guó)家自然科學(xué)基金(61300151)，江蘇省自然科學(xué)基金 (BK20130155)，江蘇省產(chǎn)學(xué)研前瞻性聯(lián)合研究項(xiàng)目(BY2013015- 02)，中央高?；究蒲袠I(yè)務(wù)費(fèi)專項(xiàng)資金資助重點(diǎn)項(xiàng)目(JUSRP 51614A)
Multi-view TSK Fuzzy System via Collaborative Learning
Funds:
The National Natural Science Foundation of China (61300151), The Natural Science Foundation of Jiangsu Province (BK20130155), The RD Frontier Grant of Jiangsu Province (BY2013015-02), The Fundamental Research Funds for the Central Universities (JUSRP51614A)
-
摘要: 傳統(tǒng)模糊系統(tǒng)建模方法本質(zhì)上是一種單視角學(xué)習(xí)模式，面向適合多視角處理的場(chǎng)景時(shí)，它們通常只能將每一視角割裂開來進(jìn)行獨(dú)立建模，這導(dǎo)致其所得系統(tǒng)泛化性能往往不令人滿意。針對(duì)此缺陷，該文探討具備多視角學(xué)習(xí)能力的模糊系統(tǒng)建模方法。為此，基于經(jīng)典的L2型TSK模糊系統(tǒng)，通過引入具備多視角學(xué)習(xí)能力的協(xié)同學(xué)習(xí)項(xiàng)，該文提出了核心的多視角TSK型模糊系統(tǒng)(MV-TSK-FS)建模方法。MV-TSK-FS不僅能有效地利用各視角不同特征構(gòu)成的獨(dú)立樣本信息，還能充分地利用各視角間由于相互關(guān)聯(lián)而存在內(nèi)在信息，以最終達(dá)到提高系統(tǒng)泛化性能的效果。在模擬數(shù)據(jù)集與真實(shí)數(shù)據(jù)集上的實(shí)驗(yàn)結(jié)果驗(yàn)證了較之于傳統(tǒng)單視角模糊建模方法該多視角模糊系統(tǒng)有著更好的泛化性和適用性。
-
關(guān)鍵詞:
- 多視角學(xué)習(xí) /
- 協(xié)同學(xué)習(xí) /
- 模糊建模 /
- TSK型模糊系統(tǒng)
Abstract: Conventional fuzzy system modeling methods essentially belong to the single-view learning modality. In multi-view-oriented data scenarios, they can only cope with each view separately, which is prone to incurring their unsatisfactory generalization performance. In response to such problem, the fuzzy system modeling method with the ability of multi-view learning is pursued. To this end, based on the classic L2 norm Takagi-Sugeno-Kang (TSK) fuzzy system, by means of the collaborative learning items qualified for multi-view learning, the core Multi-View TSK Fuzzy System (MV-TSK-FS) modeling method is presented. MV-TSK-FS can not only effectively utilize the independent components composed of the characteristics affiliated to each view, but also take full advantage of the potential information occurred by the interrelated effects among views, which eventually facilitates its relatively strong generalization ability. The experimental results performed on both synthetic and real-life datasets indicate that, compared with some traditional single-view methods, this propounded multi-view fuzzy modeling system owns preferable applicability as well as generalization. -
LI Guangxia, CHANG Kuiyu, and HOI S C H. Multiview semi-supervised learning with consensus[J]. IEEE Transactions on Knowledge and Data Engineering, 2012, 24(11): 2040-2051. doi: 10.1109/TKDE.2011.160. HONG Chaoqun, YU Jun, YOU Jane, et al. Multi-view ensemble manifold regularization for 3D object recognition [J]. Information Sciences, 2015, 320: 395-405. doi: 10.1016/ j.ins.2015.03.032. ZHANG Shunli, YU Xin, SUI Yao, et al. Object tracking with multi-view support vector machines[J]. IEEE Transactions on Multimedia, 2015, 17(3): 265-278. doi: 10.1109/TMM. 2015.2390044. 蔣亦樟, 鄧趙紅, 王駿, 等. 熵加權(quán)多視角協(xié)同劃分模糊聚類算法[J]. 軟件學(xué)報(bào), 2014, 25(10): 2293-2311. doi: 10.13328/j. cnki.jos.004510. JIANG Yizhang, DENG Zhaohong, WANG Jun, et al. Collaborative partition multi-view fuzzy clustering algorithm using entropy weighting[J]. Journal of Software, 2014, 25(10): 2293-2311. doi: 10.13328/j.cnki.jos.004510. TZORTZIS G F and LIKAS A C. Multiple view clustering using a weighted combination of exemplar-based mixture models[J]. IEEE Transactions on Neural Network, 2010, 21(12): 1925-1938. doi: 10.1109/TNN.2010.2081999. JIANG Yizhang, CHUNG Fulai, WANG Shitong, et al. Collaborative fuzzy clustering from multiple weighted views[J]. IEEE Transactions on Cybernetics, 2015, 45(4): 688-701. doi: 10.1109/TCYB.2014.2334595. MERUGU S, ROSSET S, and PERLICH C. A new multi- view regression approach with an application to customer wallet estimation[C]. Proceedings of 12th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Philadelphia, USA, 2006: 656-661. DENG Zhaohong, JIANG Yizhang, CHOI Kupsze, et al. Knowledge-leverage-based TSK fuzzy system modeling[J]. IEEE Transactions on Neural Networks and Learning Systems, 2013, 24(8): 1200-1212. doi: 10.1109/FUZZ-IEEE. 2014.6891544. STRM K J and MCAVOY T J. Intelligent control[J]. Journal of Process Control, 1992, 2(3): 115-127. 鄧趙紅, 張江濱, 蔣亦樟, 等. 基于模糊子空間聚類的〇階L2型TSK模糊系統(tǒng)[J]. 電子與信息學(xué)報(bào), 2015, 37(9): 2082-2088. doi: 10.11999/JEIT150074. DENG Zhaohong, ZHANG Jiangbin, JIANG Yizhang, et al. Fuzzy subspace clustering based zero-order L2-norm TSK fuzzy system[J]. Journal of Electronics Information Technology, 2015, 37(9): 2082-2088. TAKAGI T and SUGENO M. Fuzzy identification of systems and its application to modeling and control[J]. IEEE Transactions on Systems, Man, and Cybernetics, 1985, 15(1): 116-132. doi: 10.1109/TSMC.1985.6313399. MAMDANI E H. Application of fuzzy logic to approximate reasoning using linguistic synthesis[J]. IEEE Transactions on Computers, 1977, 26(12): 1182-1191. doi: 10.1109/TC.1977. 1674779. AZEEM M F, HANMANDLU M, and AHMAD N. Generalization of adaptive neural-fuzzy inference systems[J]. IEEE Transactions on Neural Networks, 2000, 11(6): 1332-1346. doi: 10.1109/72.883438. JIANG Yizhang, CHUNG Fulai, and WANG Shitong. Enhanced fuzzy partitions vs data randomness in FCM[J]. Journal of Intelligent and Fuzzy Systems, 2014, 27(4): 1639-1648. doi: 10.3233/IFS-141130. ITO K and NAKANO R. Optimizing support vector regression hyperparameters based on cross-validation[C]. Proceedings of the International Joint Conference on Neural Networks, Jantzen Beach, Portland, Oregon, 2003: 2077-2082. -
計(jì)量
- 文章訪問數(shù): 1562
- HTML全文瀏覽量: 160
- PDF下載量: 441
- 被引次數(shù): 0
下載:
