SOM toolbox clasificate data

Question

Is there any way to use trained neural network using "SOM toolbox" for classification of data in data set? For example I have data, I put it to the network and network tells me the type of data.

Answer 1

I do not know if you are asking for any specific SOM toolbox but I will give a general idea.

First of all it is possible to use SOM to classify data as long as you have some labelled training data, or otherwise you classify each node of the network in one specific class:

In the first case, you train your network with both inputs and desired result as features. After learning, you give the new test data to classify with only the inputs (without the desired result). The network give you back which was the best matching unit, and with this you can access to which desired result it corresponds.

In the second case you train your network in the normal way (with only the inputs as features). You classify the different network nodes. After learning, you give the new test data to classify again with only the inputs. The network give you back which was the best matching unit, and with this you can access to which class it points to.

The second case should be straightforward in any toolbox, yet the first one is not. For the first case my simple suggestion (without coding yourself) would be to train 2 networks: one with both the inputs and desired result as features, and one with only the inputs. Replace the weights discovered by the learning process of the network with only the inputs, with the ones from the inputs+result and now you can use this as normal. Get the BMU and point to the respective class.