Annexe 1 -Parameters in the SOM processing
1. net architecture
Input Unit |
Number of units in the input vector (12) |
K Units |
Number of units in the output matrix (from 9 to 49, depending on the simulation) |
K Rows |
Number of rows in the output matrix (from 3 to 7, depending on the simulation) |
K Cols |
Number of columns in the output matrix (from 3 to 7, depending on the simulation) |
K Dimension |
Output matrix dimensions (2) |
K Topology |
Output matrix space topology (Euclidean) |
N Topology_________ |
Winner unit neighbourhood space topology (square)___________________________________ |
2. parameters
N function |
Parameter defining the function to update the units connections in the WU neighbourhood |
Alpha Max_________ |
Maximum width for the Nfunction (1)____________________________________________ |
Alpha Min__________ |
Minimum width for the Nfunction (0)_______________________________________________ |
Alpha Inc___________ |
Factor reducing Alpha Max in each epoch (0.01)______________________________________ |
Set Weight___________ |
Maximum weight value during the initialisation_________________________________________ |
Alpha W Func______ |
Input/output weights correction function (constant)________________________________________ |
Alpha W Max_______ |
Initial value of weight correction factor (0.1)_______________________________________________ |
Alpha WMin_______ |
Minimum value of weight correction factor (0)_______________________________________ |
Alpha W Inc________ |
Decreasing amount of the weight correction factor (0.001)_______________________________ |
Epochs |
The epochs number for an experiment is automatically calculated by this formula: A A AlphaMax - AlphaMin AlphaInc |
3. Input record
I Patterns ∣ Number of records in the input sample (2703)
19