Relate data points to clusters after clustering with GNG

[Rbasarat]

Hi,

I am doing some research where I cluster data with the GNG. For this research I need to know which data points are in which clusters. Is it possible to check which data point is in which cluster after training the GNG?

[itdxer]

Hi @Rbasarat,

Current implementation doesn't store this information, but it can be calculated with a few lines of code. You can get location of the nodes and find closest neuron per each sample individually with l2 distance.

from neupy import algorithms
import numpy as np
from scipy.spatial import distance_matrix

gng = algorithms.GrowingNeuralGas(...)
gng.train(X_samples)

# The main logic
gng_nodes = np.concatenate([node.weight for node in gng.graph.nodes])
closest_node_index_per_sample = np.argmin(distance_matrix(X_samples, gng_nodes, p=2), axis=1)

(I haven't check the code, so I'm not 100% that it won't fail, but at least I hope the idea will be clear)