Loading the recall file |
In order to recall the network you need to load the recall file.
To load the training file, press the Open button and browse to find the required file.
Alternatively, select Load Recall File from the Recall menu and browse to find the required file.
The file can have any extension, but the convention is for recall files to have the extension *.rcl and be of the required format.
Viewing the recall file |
You can press the View button to view your recall file, or select View Recall File from the Recall menu.
Recalling the network |
Once the recall file has been loaded, press the Recall button, or select Recall from the Recall menu.
The recall output now needs to be saved to a file.
Saving the output file |
Once the recall button has been pressed, you need to save the output coordinates.
To specify the output file, press the Open button and give the file a name, e.g. iris_recall.cod.
Alternatively, select Load Output File from the Recall menu and specify the file.
The file can have any extension, but the convention is for output coordinates files to have the extension *.cod.
Viewing the output file |
You can press the View button to view your output file, or select View Output File from the Recall menu.
What you see are the inputs from the recall file and then the coordinates of the node of the map that each recall row activated.
Below is an example output generated by training of a Kohonen network.
What you see are the inputs from the recall file and then the coordinates of the node of the map that each recall row activated.
[FormatVersion = 1.0] /* This is a standard version number. Any value larger than 1.0 will not work*/
[DataSet] /*Details of the data set*/
[Rows = 75]
[Inputs = 4]
[ExpectedOutputs = 2]
[GeneratedOutputs = 0]
[Error = 0]
[Data] /*the input values for each row from the recall data file, with the coordinates of the node they activated*/
0.171429 | 0.368421 | 0.076923 | 0.041667 | 0 | 4 |
0.171429 | 0.578947 | 0.076923 | 0.125 | 0 | 4 |
0.628571 | 0.368421 | 0.615385 | 0.541667 | 0 | 3 |
0.685714 | 0.263158 | 0.692308 | 0.541667 | 1 | 3 |
0.657143 | 0.368421 | 0.730769 | 0.666667 | 1 | 2 |
0.457143 | 0.315789 | 0.634615 | 0.583333 | 3 | 3 |
0.371429 | 0.157895 | 0.442308 | 0.375 | 3 | 4 |
0.314286 | 0.052632 | 0.5 | 0.416667 | 3 | 4 |
0.314286 | 0.052632 | 0.480769 | 0.375 | 3 | 4 |
0.4 | 0.210526 | 0.519231 | 0.458333 | 3 | 4 |
0.457143 | 0.210526 | 0.75 | 0.625 | 3 | 2 |
0.285714 | 0.368421 | 0.634615 | 0.583333 | 1 | 4 |
[~DataSet]
Labeling the output coordinates |
Once you have saved the recall output coordinates, press the Label button, or alternatively select Label from the Label menu.
When you label the network you create a file that contains the label from the map file corresponding to the coordinates of the node on the map that that row in the recall data activated the most.
For example:
Map Coordinate File
[FormatVersion = 1.0] [DataSet]
[Data]
[~DataSet] |
Recall Output File
[FormatVersion = 1.0] [DataSet]
[Data]
[~DataSet] |
Network Label File
[FormatVersion = 1.0] [DataSet]
[Data]
[~DataSet] |
After labeling the network you should save the file by selecting Save As Label File from the Label menu.