Researches into Neural Network Architecture
dc.contributor.advisor | Sánchez-Sinencio, Edgar | |
dc.creator | Ewell, James J. III | |
dc.date.accessioned | 2022-04-01T16:03:53Z | |
dc.date.available | 2022-04-01T16:03:53Z | |
dc.date.issued | 1990 | |
dc.identifier.uri | https://hdl.handle.net/1969.1/CAPSTONE-RogersS_1977 | |
dc.description | Program year: 1989/1990 | en |
dc.description | Digitized from print original stored in HDR | en |
dc.description.abstract | New insight into the choice of the number of hidden nodes for three layered neural networks is presented. I have conceived a method to reduce the total complexity and size of a network by testing for linear dependence of nodal relationships. Limitations imposed by choice of the training algorithm and training set is introduced. A discussion of training and its effects on neural net performance completes the discussion of neural networking. These evaluation procedures are very important for the physical construction of neural networks. | en |
dc.format.extent | 24 pages | en |
dc.format.medium | electronic | en |
dc.format.mimetype | application/pdf | |
dc.subject | neural networks | en |
dc.subject | nodal relationships | en |
dc.subject | training algorithm | en |
dc.subject | linear dependence | en |
dc.title | Researches into Neural Network Architecture | en |
dc.type | Thesis | en |
thesis.degree.department | Electrical Engineering | en |
thesis.degree.grantor | University Undergraduate Fellow | en |
thesis.degree.level | Undergraduate | en |
dc.type.material | text | en |