| Creator |
|
|
|
|
|
| Language |
|
| Publisher |
|
|
|
| Date |
|
| Source Title |
|
| Vol |
|
| Publication Type |
|
| Access Rights |
|
| Related DOI |
|
| Related DOI |
|
| Related URI |
|
| Related URI |
|
| Related HDL |
|
| Relation |
|
| Abstract |
The concept of domains of recognition is introduced for three-layered neural networks. The domain lies in the input space and can be represented using connection weights and thresholds of the network.... We propose a learning method of the network so as to enlarge the domain of recognition by extending its range mapped into the hidden space and by minimizing the slope of affine transforms in the mapping. Based on the method, we introduce a cost function whose minimization process gives a learning algorithm of the network. A land cover classification problem has been considered in our simulation.show more
|
Export Mendeley
trcs172.ps
trcs172