| Scenario |
Type |
Detail |
| Basic parameters of the network |
Basic Path |
The user must insert some parameters of the network, that are necessary to create the net correctly. Basic parameters are :
- number of hidden layers (integer always > 0) - hidden because there is always one input and output layer
- number of nodes in each hidden layer (integer or integers divided by semicolon) - as many values as the number of hidden layers is
- stopping parameter (in which case the layer is going to stop with learning) - the only one condition is after the selection used
- number of learning cycles (integer)
- timeout in miliseconds (integer)
- total error of the network in % (double)
Every parameter is checked for its correctness before leaving the dialog. |
| Implicit advanced parameters |
Basic Path |
If the user do not want to fill in advanced parameters, than these are used :
- learning parameter = 0,5
- learning moment = 0,05
- slope of the sigmoid function = 1
- initial weights and thresholds = 0,2
|
| Advanced parameters of the network |
Alternate |
The user is able to define some advanced parameters of the network
- learning parameter (double)
- learning moment (double)
- the slope of sigmoid function (double)
- initial maximal values of thresholds and weights of each node (initial value is a random number in absolute value not higher than this number)
Filled numbers are controlled for their correctness. |