References in periodicals archive ?
Computationally, the assumption that the tasks faced by the learner are related to one another represents a powerful bias at the metalevel, ensuring that metalearning algorithms do not run afoul of the NFL theorems or the LCGP.
LCGP reduces the number of levels in the planning graph by allowing more parallelism; mutex actions can appear in the same level if there is a feasible serialization.
New Results about LCGP, A Least Committed Graphplan.