Information Processing Features Can Detect Behavioral Regimes of Dynamical Systems

Figure 2

Predictive power of the optimal set of information features as function of (thick blue line with round markers). For each the best set of information features is listed: S means “synergy” and I means mutual information, followed by a bitmask of which neighbor cell’s initial state is included in the feature (middle bit is the cell itself). For example, I010 indicates the “memory” feature , and S111 denotes . The thin black line with error bars is the “base line” predictability for features, obtained by randomizing the pairing of 256 information features with their class identifier. The error bar indicates the 95% confidence interval of the distribution of predictive power under the null-hypothesis of zero correlation between information features and class identifier. Finally, the small black markers indicate the predictive powers of all other information feature sets.

(a)

(b)

(c)