Figure 9.

The success rate of the three learners in a low-entropy (highly ordered) network: a dominant trajectory among the elements in the network occurs with a high frequency, allowing high-accuracy prediction by model-constructing learners, thus giving a significant advantage to chaining and CL. CL has a significant advantage over chaining when the training set is shortened from 4000 to 2000 characters, (a) and (b), respectively. See related TP matrix in the electronic supplementary material, table T8.