The B-Spline Editor can be used to define a 20-sim B-Spline network. The editor is opened when you try to edit this submodel (using the Go Down command).
|•||Network Name: this is the local name of the submodel representing the B-Spline network (this option is not yet supported).|
|•||Order: Enter the order of the B-Splines that must be used.|
|•||Learn at Each Sample: Select this option if you want the network to adapt the weights after every simulation step|
|•||Learning Rate: Specify the learning rate that should be used during training. If you do not want the network to learn, enter a zero value here.|
|•||Apply Regularization: This option is not yet supported.|
|•||Load Weights at Start of Simulation: Select this option, if you want to use predefined weights stored on file (saved in a previous run). Before each simulation run, you will be asked to enter the filename of this weights file.|
|•||Save Weights at End of Simulation: Select this option, if you want to store the weights on file. After each simulation run, you will be asked to enter the filename of this weights file.|
|•||Network is Discrete: Select this option if the B-spline network is connected with discrete-time models. Deselect this option if the network is connected with continuous models.|
The tabs shown in the editor is used to specify the inputs of the network. Each tab shows the specific settings of that input.
|•||Add Input: Add a new input to the network. A new tab will be added with default settings.|
|•||Delete Input: Delete the input defined in the selected tab (the tab that is in front).|
Each input is associated with a certain number of splines. To give the user more flexibility, each input can be divided in certain regions that each have their own density of splines.
|•||Add: Add a new region of splines.|
|•||Delete: Delete the selected region of splines.|
|•||Split: Split the selected region into two regions.|
Each region can be edited. Just select it from the Regions list and use the following options:
|•||Upper / Lower: The upper and lower bound of a region.|
|•||Number of Splines: The number of splines. The more splines you choose, the more accurate will the network output be, for this region of inputs.|