Fig. 4: MPL searches more efficiently than other models.
From: Symbolic metaprogram search improves learning efficiency and explains rule learning in humans
A Loge posterior of the best solution discovered by a given loge search step per function (n = 100 functions; thick = mean) per model with a fixed training set of 10 input/output examples per function. (B) and (C) follow Fig. 3A, B, respectively, with 5K search steps per trial.
