Question

Which of the following is incorrect about Self-Organizing Maps?

(a) Clustering by SOMs is in principle similar to the k-means method

(b) It doesn’t involve neural networks

(c) The data points are initially assigned to the nodes at random

(d) It starts by defining a number of nodes

This question was posed to me in an internship interview.

Origin of the question is Microarray-Based Approaches in section Functional Genomics & Proteomics of Bioinformatics

Answer 1

The correct option is (b) It doesn’t involve neural networks

Explanation: This pattern recognition algorithm employs neural networks. The distance between the input data points and the centroids are calculated. The data points are successively adjusted among the nodes, and their distances to the centroids are recalculated. After many iterations, a stabilized clustering pattern are reached with the minimum distances of the data points to the centroids. The differences between SOM and k-means are that, in SOM, the nodes are not treated as isolated entities, but as connected to other nodes.