>> net.layers{:}
ans =
Neural Network Layer
name: 'Layer'
dimensions: [8 8]
distanceFcn: 'linkdist'
distanceParam: (none)
distances: [64x64 double]
initFcn: 'initwb'
netInputFcn: 'netsum'
netInputParam: (none)
positions: [2x64 double]
range: [64x2 double]
size: 64
topologyFcn: 'hextop'
transferFcn: 'compet'
transferParam: (none)
userdata: (your custom info)
Assuming you are using the Matlab version of selforgmap, this is where you can find the neighbourhood connections, which are fixed by the starting topology so do not change at all during the learning of the network. The positions property gives an array of the x-y locations of the nodes as defined by the topology. You can also access the topology outwith the SOM using e.g.
doc hextop
if you are using the hexagonal toplogy (the default), since it is unaffected by the running of the SOM.
I did always find it very hard to find or remember which properties contained the information I wanted when I used this myself. I ended up writing my own version of an SOM in the end because I wanted more flexibility.
