A Competitive Wavelet Layer for Pattern Clustering

作者: Roberto Kawakami Harrop Galvão , Takashi Yoneyama

DOI: 10.21528/CBRN1999-021

关键词: Pattern recognitionCluster (physics)Layer (object-oriented design)WaveletMaxima and minimaRepresentation (mathematics)Basis (linear algebra)Wavelet packet decompositionMachine learningMathematicsArtificial intelligenceCluster analysis

摘要: A competitive “wavelet layer” is proposed for pattern clustering. It exploits the representation capabilities of adaptive wavelets to generate template approximations each cluster data. brief review wavelet representations, as well some insight into local minima problems, provided. The method illustrated by a simple clustering problem, in which step responses dynamic systems are discriminated with basis on presence parasitic oscillations. results suggest that layer exhibits superior performance than conventional neural layers when patterns exhibit low signal-to-noise ratio.

参考文章(8)
Gilbert Strang, Truong Nguyen, Wavelets and filter banks ,(1996)
Harold H Szu, Brian A Telfer, Shubha L Kadambe, Neural network adaptive wavelets for signal representation and classification Optical Engineering. ,vol. 31, pp. 1907- 1916 ,(1992) , 10.1117/12.59918
H. Dickhaus, H. Heinrich, Classifying biosignals with wavelet networks [a method for noninvasive diagnosis] IEEE Engineering in Medicine and Biology Magazine. ,vol. 15, pp. 103- 111 ,(1996) , 10.1109/51.537066
Ingrid Daubechies, Ten Lectures on Wavelets ,(1992)
H. Szu, Intelligent neural networks using smart wavelets preprocessing international conference on tools with artificial intelligence. pp. 348- 349 ,(1994) , 10.1109/TAI.1994.346388
Q. Zhang, A. Benveniste, Wavelet networks IEEE Transactions on Neural Networks. ,vol. 3, pp. 889- 898 ,(1992) , 10.1109/72.165591