RSS Fingerprint indoor localization using neural and SVR algorithm about smile HOT 10 CLOSED

The first phase of RBF networks is just k-means. If you need only that, why not just use k-means?

from smile.

Kmeans need predefined number of cluster and its one from disadvantages.
Affinity Propogation used to cluster all data after that I use rbf neural for train all cluster alone.
Ii implement all one from components in graph above .
My Qz how I can integrate this part?
Any code integrate between cluster and train all one from cluster with rbf nn will be helpful

from smile.

K-means inside rbf networks still needs the number of clusters. The trainer class has a method setNumCenters(int) to set it.

from smile.

I can use just one cluster algorithm
I use affinity propagation for cluster data at first .
and I can use same algorithm in rbf with predefined centroid to build gaussian
thats correct?

from smile.

Yes, you can use predefined centroids. When creating a RBFNetwork instance, you can provide them in the constructor.

from smile.

If I train all one from cluster in rbf neural network .
in this case all cluster have one centroid?

from smile.

I don't understand your question. It is not valid english. sorry.

from smile.

I ask about
after use cluster algorithm for data
Each one will have just one centroid in this case?

from smile.

For many partitioning clustering algorithms (e.g. k-means), each cluster has just one centroid.

from smile.

Thank you

from smile.