Skip to main content

Table 1 SOM parameters together with allowed ranges.

From: Conformational and functional analysis of molecular dynamics trajectories by Self-Organising Maps

SOM parameter Range
Map size [100,400]
Lattice type {hexagonal, rectangular}
Shape {sheet, cylinder, toroid}
Learning algorithm {batch, sequential}
Neighbour function {gaussian, bubble, ep}
Alpha type {inverse, linear, power}
Radius {1, 2, 3}
Training length [1000,5000]
Starting alpha [0.01, 0.09]
  1. Map size (number of neurons), Lattice type, and Shape define the topological characteristics of the map. The other parameters are associated to the learning process: Learning algorithm, Neighbour function, Alpha type, Radius (neighbourhoods' kernel), Training length (number of learning cycles or epochs), Starting alpha (parameter defining the initial learning rate).