Investigations in neural network learning using geometrical approach
In this thesis several methods to control the geometrical expansion process are proposed and analyzed. Our approaches turn out to gain faster training speed with satisfactory accuracy. We discuss problems both with Boolean inputs and with real-valued inputs respectively.
Saved in:
主要作者: | |
---|---|
其他作者: | |
格式: | Theses and Dissertations |
出版: |
2008
|
主題: | |
在線閱讀: | https://hdl.handle.net/10356/2517 |
標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|