In this thesis we extensively experiment with two ART (adaptive resonance theory) architectures called Gaussian ARTMAP (GAM) and Distributed Gaussian ARTMAP (dGAM). Both of these classifiers have been successfully used in the past on a variety of applications. One of our contributions in this thesis is extensively experiments with the GAM and dGAM network parameters and appropriately identifying ranges for these parameters for which these architectures attain good performance (good classification performance and small network size). Furthermore, we have implemented novel modifications of these architectures, called semi-supervised GAM and dGAM architectures. Semi-supervision is a concept that has been used effectively before with the FAM and EAM architectures and in this thesis we are answering the question of whether semi-supervision has the same beneficial effect on the GAM architectures too. Finally, we compared the performance of GAM, dGAM, EAM, FAM and their semi-supervised versions on a number of datasets (simulated and real datasets). These experiments allowed us to draw appropriate conclusions regarding the comparative performance of these architectures.
| Identifer | oai:union.ndltd.org:ucf.edu/oai:stars.library.ucf.edu:etd-1293 |
| Date | 01 January 2005 |
| Creators | Chalasani, Roopa |
| Publisher | STARS |
| Source Sets | University of Central Florida |
| Language | English |
| Detected Language | English |
| Type | text |
| Format | application/pdf |
| Source | Electronic Theses and Dissertations |
Page generated in 0.0023 seconds