Is there a way to create clusters of equal size?

[mpati005]

If you take a dataset of 1000 objects and create clusters (ref: https://github.com/JustGlowing/minisom/blob/master/examples/Clustering.ipynb), you end up getting clusters of different sizes.

Further, using
# each neuron represents a cluster
winner_coordinates = np.array([sog.winner(x) for x in data]).T
# with np.ravel_multi_index we convert the bidimensional
# coordinates to a single dimensional index
cluster_index = np.ravel_multi_index(winner_coordinates, sog_shape)****

each cluster has 99, 43, 94, 74, 121 and so on number of objects.

Is there a way to create clusters that have same number of objects?

[JustGlowing]

hi @mpati005, you can try changing the number of clusters so that the samples are almost evenly distributed. Obviously there are more clever approaches but the outcomes highly depend on the data you're using.

[mpati005]

Hi @JustGlowing

Thank you.

I have considered another option which is to try to change the SOM but regardless of any combination of (n*m) SOM, I am yet to see equal number of objects associated with all neurons.

Having said that, and adding to what you suggest, it might be better to create clusters based on the data points associated with the neurons first and then alter the clusters to create equal sizes.
Will update if I see something else.