In MATLAB, clustering data using kmeans can be achieved as shown below:
L = kmeans(X,k,Name,Value) where L is cluster indices which is for each data point.
It implies that is if I have 307 data points I'm to have a 307 x 1 array(L) which is the index for each data point.
However, while using SOM for clustering I discovered to get the index you use the code snippet below:
net = selforgmap([dimension1 dimension2]);
% Train the Network
[net,tr] = train(net,X);
%get indices
L = vec2ind(net(X))';
for a Network with 5 x 5 dimension:
it returns L which is an array with the dimesion 25 x 1 instead of 307 x 1 for a Network with 10 x 10 dimension:
it returns L which is an array with the dimesion 100 x 1 instead of 307 x 1
What am I doing wrong???
or to simply put, how do I compute the class vectors of each of the training inputs ?
Dear Dr. Alabi,
I tried very hard to come to a conclusion to your problem and it seems it is very simple.
The following must be done:
net = newsom([0 0],[5 5]);
[net,trn,Y] = train(net,double(L'));
You get Y 25*307 where each column should be either 1 or 0 at the end if you multiply each row by the cluster number (1 to 25) and then add all rows together you get new clustered data into 25 clusters for the 307. This means you have 1*307 each has a value between 1 and 25 representing the cluster number.
Best regards
Thanks Dr. Awad for your response sir. What you have there seems right sir. I will try that out and get back to you sir.
However another question for you sir, Does it necessarily mean that the clustered data must have 25 clusters? or are we indirectly saying it can only have maximum of 25 clusters?
Dear Dr. Alabi,
Your question is very important. You may have a a network of 10X10 but not all are used. To answer your question you should read my paper :
Chapter Segmentation of Satellite Images Using Self-Organizing Maps
because sometime in case of SOMs clusters you may have oversegmentation and you have to use thresholds to remove redundant clusters or combine them.
Best regards
Dr. Awad,
I have tried this sir:
net = selforgmap([dimension1 dimension2]);
% Train the Network
[net,tr,Y] = train(net,double(X));
but Y still has the dimension 100 x 1 instead of 307 x 1
Note sir, X has 26 attributes(columns) and 307 tuples(rows).
Dear Jesujoba,
Most probably you should input X transposed to the train function (being the first dimension of X the number of features and the second dimension, the number of training samples). Here is a Matlab example code for training a 7x7 SOM network:
>> X = rand([26 307]);
>> net = selforgmap([7 7]);
>> net = train(net,X);
>> y = net(X);
>> classes = vec2ind(y);
>> size(classes)
ans =
1 307
Where classes is a vector of 1x307, where each of the 307 training samples has been assigned to one of the 49 units in the network.
Hope this helps.
That works fine Dr Carlos Sampedro Pérez. Kudos to you sir
Dear Dr. Alabi,
I do not agree with the solution provided by Carlos because it depends on using random numbers. The procedure is good, however it should not be random. it should work as follow:
net = newsom([0 0],[7 7]);
net.inputs{1}.size = 26;
[net,trn,Y] = train(net,double(X));
%Where is is your input with N=26 bands or nodes; i.e. sizeof(X) = N,307
classes = vec2ind(Y);
Try it with confidence
Best regards
Dear M. Awad,
The purpose of my previous answer was just for providing a little example. Obviously, the random training samples of my previous answer have to be substituted by the actual training samples of the corresponding problem.
Kind regards,
Carlos.
- Badges
- Science topic
- Similar topics
- Computer Science
- Data Mining