Please use this identifier to cite or link to this item:
Title: Hierarchical clustering using K-Iterations Fast Learning Artificial Neural Networks (KFLANN)
Authors: Wong, Lai Ping
Keywords: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
Issue Date: 2007
Source: Wong, L. P. (2007). Hierarchical clustering using K-Iterations Fast Learning Artificial Neural Networks (KFLANN). Doctoral thesis, Nanyang Technological University, Singapore.
Abstract: Hierarchical clustering using hybrid learning model of KFLANN and Multilayer Perceptron with Backpropagation learning algorithm (MLP-BP) is proposed to address high dimensional classification problems. K-Iterations Fast Learning Artificial Neural Network (KFLANN) was enhanced to tackle the sensitivity of clustering against Data Presentation Sequence. Number of cluster is not required prior clustering process for KFLANN clustering algorithm. Data driven scheme is used to define network parameters and only small number of iterations is needed for the algorithm to converge. The KFLANN tends to cumbersome when feature dimensionality is large. HieFLANN and HieFLANN-BP were proposed to avoid this cumbersome. Hierarchical network made up of KFLANN (HieFLANN) was developed to address the limitation of KFLANN in handling large dimensionality problem set. HieFLANN performs clustering and data transformation within a single model. Data transformation adopts canonical covariance concept. HieFLANN only perform classical clustering on a given problem set, thus it lacks of generalization ability. HieFLANN-BP with hybrid learning model as its subunits was build to tackle this issue. It inherits generalization capability from the MLP-BP. Performance of a learning system tends to drop when portion of irrelevant information increases. Feature selection scheme based on purity and relevance (PURE) was proposed to filter irrelevant information.
DOI: 10.32657/10356/2530
Schools: School of Computer Engineering 
Rights: Nanyang Technological University
Fulltext Permission: open
Fulltext Availability: With Fulltext
Appears in Collections:SCSE Theses

Files in This Item:
File Description SizeFormat 
SCE-THESES_276.pdf1.32 MBAdobe PDFThumbnail

Page view(s) 50

Updated on Jun 12, 2024

Download(s) 20

Updated on Jun 12, 2024

Google ScholarTM




Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.