During the learning process, neurons that are topographically close in the array will activate each other to learn something from the same input vector. This results in a smoothing effect on the weight vectors of neurons. Thus, these weight vectors tend to approximate the probability density function of the input vector. Therefore, the visualization of elements of these vectors for different input variables is convenient to understand the contribution of each input variable with respect to the clusters on the trained SOM. Therefore, to analyze the contribution of variables to cluster structures of the trained SOM, each input variable (component) calculated during the training process can be visualized in each neuron on the trained SOM in grayscale.
Was this article helpful?