A method to integrate top down and bottom up induction for augmentation of incomplete theories

The incomplete theory problem has been of great interest both in explanation-based learning and more recently in inductive logic programming. Earlier methods used either top-down or bottom-up induction, and both these induction strategies included some undesired restrictions on the hypothesis space. To overcome these limitations a method is presented where the different induction strategies are completely integrated, which not only eliminates the restrictions that they imply when separated but also makes possible the use of an overall preference bias.<>