Analysis of information gain and Kolmogorov complexity for structural evaluation of cellular automata configurations
Mohammad Ali Javaheri Javid, Tim Blackwell, Robert M. Zimmer, Mohammad Majid al‐Rifaie · Connection Science · 2016
Shannon entropy fails to discriminate structurally different patterns in two-dimensional images. We have adapted information gain measure and Kolmogorov complexity to overcome the shortcomings of entropy as a measure of image structure. The measures are customised to robustly quantify the complexity of images resulting from multi-state cellular automata (CA). Experiments with a two-dimensional multi-state cellular automaton demonstrate that these measures are able to predict some of the structural characteristics, symmetry and orientation of CA generated patterns.