DSpace@nitr >
National Institue of Technology- Rourkela >
Thesis (MTech by Research) >

Please use this identifier to cite or link to this item: http://hdl.handle.net/2080/669

Full metadata record

DC FieldValueLanguage
contributor.authorVamsi Krishna, T-
contributor.authorSatapathy, J K (Guide)-
identifier.citationDevelopment Of Novel Neuro-Fuzzy Techniques For Adaptive Systems, Thesis submitted in partial fulfillment of the requirements for the degree of Master of Technology by Research, to National Institute of Technology, Rourkelaen
descriptionCopyright for the article belongs to National Institute of Technology, Rourkelaen
description.abstractNovel approaches for designing adaptive schemes based on neuro-fuzzy platform have been developed. Two kinds of adaptive schemes namely, adaptive equalization and system identification are implemented using the developed proposed techniques. The Radial basis function (RBF) equalizer is chosen as a case study for adaptive equalization of the digital communication channels. An efficient method for reducing the centers of a RBF equalizer based on eigenvalue analysis is presented. The efficiency of the method is further verified for RBF equalizers with decision feedback for tackling channels with overlapping channel states. A comparative study between the proposed center reduction technique and other center reduction techniques for the RBF equalizer is discussed. In another breakthrough a parallel interpretation of the ANFIS (adaptive network based fuzzy inference systems) architecture is proposed. This approach helps to investigate the role of the fuzzy inference part and the sub-filter part of the ANFIS separately. The parallel interpretation of the ANFIS redefines the opinion reserved for the fuzzy inference system, thereby allowing it to be considered as a fuzzy weighted sub-filter network, with the weighting functions and the subfilter units arranged parallely. This approach motivated in developing many novel schemes for designing adaptive systems with application to system identification problems. Finally, the limitations of the ANFIS architecture are discussed. These limitations are exploited to develop neuro-fuzzy models similar to the ANFIS with the objective of reducing the number of parameters in comparison to the ANFIS. The developed neuro-fuzzy models are compared to the ANFIS in terms of the time required for learning and number of parameters to be adapted.en
format.extent2670950 bytes-
publisherNational Institute of Technology, Rourkelaen
titleDevelopment Of Novel Neuro-Fuzzy Techniques For Adaptive Systemsen
Appears in Collections:Thesis (MTech by Research)

Files in This Item:

File Description SizeFormat
thesis-final,,,,,.pdf2608KbAdobe PDFView/Open

Show simple item record

All items in DSpace are protected by copyright, with all rights reserved.


Powered by DSpace Feedback