Neural Network Clustering: Querying a neuron hit on a Self organising Map
5 views (last 30 days)
Show older comments
Is there a command within the Neural Networks toolbox to query a neuron on a SOM?
I created a input training matrix of 9 animals with 7 features per animal(eg. size, legs, wings, etc). I tried to use a 2x2 SOM to classify these animals. The SOM suggested that Neuron1=2 hits, Neuron2=1 hit, Neuron3=4 hits, Neuron4=2 hits.
I would like to interrogate each neuron to find out which of the 7 vectors mentioned triggered it (eg. which two animals triggered Neuron1) so the map is more representative of which vectors are associated with various neurons.
Any help will be deeply appreciated.
PS: I've used the nctool command to generate the SOM.
Thanks. John.
0 Comments
Accepted Answer
Greg Heath
on 22 Apr 2013
Kohonen created his SOM to be used as data a visualization aid; NOT a classifier.
He developed the LVQ algorithms for classification. I'm pretty sure they are initialized with SOM.
Check with Kohonen's books and comp.ai.neural-nets.
How many nodes do you have on your map? If you want to separate the 9 animals on the map, try increasing the dimensions on the map.
Once you have a map can't you introduce just one animal to see where it falls?
Better yet: Use LVQ that's what Kohonen designed it for.
Hope this helps.
Thank you for formally accepting my answer.
Greg
0 Comments
More Answers (0)
See Also
Categories
Find more on Function Approximation, Clustering, and Control in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!