الفهرس 
AcknowledgementOnly 14 pages are availabe for public view
 |  | from | 179 |  |  |
| | | from | 179 | | |
Abstract
The statistical approach to pattern recognition is among the early approaches applied Id of research. Parametric statistical classifiers design techniques have been ly studied, in general, and Gaussian classifiers, in particular, due to its tractability [10]. However, some assumptions inherent in the design of the classifier result in suboptimal classifier [9]. 33], it was demonstrated, both theoretically and experimentally, that a neural pattern classifier generates the empirical distribution of the sample data which to train the network. This thesis takes advantage of this fact and improves a classifier using an isomorphic sigma-pi neural network [9]. dy contains four chapters, and three appendices. Their contents are as follows:
I: Presents different approaches to pattern recognition. Special attention is f to classifiers designed using the decision theoretic approach such as neural s classifiers and traditional statistical classifiers. Both of these classifiers are in more detail.
D: Presents the main drawbacks of the basic parametric Gaussian classifier method for mapping it to a Gaussian isomorphic neural network (GIN) and ’. g the GIN back to a Gaussian classifier. Backpropagation learning is reviewed modification is suggested to overcome network paralysis. The algorithms used in cation are stated along with their data structures, storage and time complexities.
r ID: The hybrid statistical neural network classifier proposed in chapter 11 is using a generated multivariate normal distributed data and two well known data ’(The Fisher’s iris data, ’and the british towns data). Also, a detailed design of a : target recognition system is presented including a review on sonar, data •sition, feature extraction and results.
the present work and discusses some possible directions for br>Appendix A: Presents the Gaussian elimination algorithm for evaluating the and the determinant and the inverse of a matrix. Also, unstable or ill-conditioned systems are
Appendix B; Presents Linear prediction and a special attention is given to the all-pole model
aapenixnd C: Lists of Fisher’s iris data, british towns data and sonar data are given.