A glimpse into the Self-Organizing Maps (SOM)

Self-Organizing Maps (SOM) are neural structures proposed for the first time by the computer scientist T. Kohonen in the late 1980s (that’s why they are also known as Kohonen Networks). Their peculiarities are the ability to auto-cluster data according to the topological features of the samples and the approach to the learning process. Contrary to methods like Gaussian Mixtures or K-Means, a SOM learns through a competitive learning process. In other words, the model tries to specialize its neurons so to be able to produce a response only for a particular pattern family (it can also be a single input sample representing a family, like a handwritten letter). Let’s consider a dataset containing N p-dimensional samples, a suitable SOM is a matrix (other shapes, like toroids, are also possible) containing (K × L) receptors and each of them is made up of p synaptic weights. The resulting structure is a tridimensional matrix W […]