AI*IA 1991: Trends in Artificial Intelligence pp 238-247 | Cite as
Extending inverse resolution to build up abstractions
Abstract
Abstraction is a conceptual framework potentially unifying and integrating different methodologies developed in machine learning and problem solving. According to the common understanding of the term, abstraction is a mapping between languages, and, then, in many learning tasks it can be so formulated. On the contrary, in this paper abstraction is defined as a mapping between models, implicitly extending Tenenberg's restricted predicate mapping in order to include more complex abstraction schemes and to cope with the problem of inconsistency in the abstract model.
In our proposal, the abstraction mapping is axiomatized by means of a theory TA, defining the semantics of the relations in the abstract model starting from the ones in the ground model. Therefore, this form of abstraction is called semantic and must be evaluated using a deductive mechanism instead of a purely syntactic rewriting. Afterwards, a restricted class of semantic abstraction (CP-Abstraction) is characterized: it has the property of preserving concept instances, with respect to a given model, and the more-general-than relation between formulas. CP-Abstraction fits in the paradigm of inverse resolution, already proposed as a framework for constructive learning and a restricted form of absorption rule is introduced to compute it.
Inverse resolution, in its original formulation, does not allow important abstraction types, such as the definition of a compound object starting from its parts, to be defined. A new operation, called term abstraction, is then introduced.
Preview
Unable to display preview. Download preview PDF.
References
- [1]A. Newell, H. Simon: Human Problem Solving, Prentice-Hall, (Englewood Cliffs, NJ, 1972).Google Scholar
- [2]D. Plaisted: “Theorem Proving with Abstraction”, Artificial Intelligence, 16, 47–108 (1981).Google Scholar
- [3]J. Tenenberg: “Preserving Consistency across Abstraction Mappings”, Proc IJCAI-87, (Milano, Italy, 1987), pp. 1011–1014.Google Scholar
- [4]C. Knoblock: “Abstractiong the Tower of Hanoi”, Working notes of the AGAA-90 Workshop, (Boston, MA, 1990), pp. 13–23.Google Scholar
- [5]C. Knoblock: “Learning Hierarchies of Abstraction Spaces”, Proc. 6th Int. Workshop on Machine Learning (Ithaca, NY, 1989).Google Scholar
- [6]G. Drastal, G. Czako, S. Raatz: “Induction in an Abstraction Space”, Proc. IJCAI-89, (Detroit, MI, 1989), pp. 708–712.Google Scholar
- [7]I. Mozetic, “Abstractions in Model-Based Diagnosis”, Working notes of the AGAA-90 Workshop, (Boston, MA, 1990), pp. 64–75.Google Scholar
- [8]A.Giordana, L.Saitta, “Abstraction: a General Framework for Learning”, Working notes of the AGAA-90 Workshop, (Boston, MA, 1990), pp. 245–256.Google Scholar
- [9]J. Carbonell, C. Knoblock, S. Minton: “PRODIGY: An Integrated Architecture for Planning and Learning”, Report CMU-CS-89-189, Carnegie-Mellon Univ. (Pittsbugh, PA), (1989).Google Scholar
- [10]R.Stepp, R. Michalski, “Conceptual Clustering, Inventing Goal-Oriented Classifications of Structured Objects”, in Michalski R., Carbonell J. & Mitchell T.(Eds) Machine Learning: An AI Approach, Vol. II. Los Altos, CA: Morgan Kaufmann (1985), pp. 471–498.Google Scholar
- [11]P.Utgoff, “Shift of Bias For Inductive Concept Learning”, in Michalski R., Carbonell J. & Mitchell T.(Eds) Machine Learning: An AI Approach, Vol. II. Los Altos, CA: Morgan Kaufmann (1985), pp. 107–148.Google Scholar
- [12]S.Muggleton, W.Buntine, “Machine Invention of First-Order Predicates by Inverting Resolution”, Proc. Fifth Int. Conf. on Machine Learning, (Ann Arbor, MI, 1988), pp.339–352.Google Scholar
- [13]C.Rouveirol, J.F. Puget,“Beyond Inversion of Resolution”, Proc. Seventh Int. Conf. on Machine Learning, (Austin, TE, 1990). pp. 122–131.Google Scholar
- [14]W. Buntine: “Generalized Subsumption and its Applications to Induction and Redundancy”, Artificial Intelligence, 36, 149–176 (1988).Google Scholar
- [15]G.Plotkin, “Automatic Methods of Inductive Inference”, Doctoral thesis, Edimburg University.Google Scholar
- [16]S. Muggleton, “Duce, an Oracle Based Approach to Constructive Induction,” in Proc IJCAI-87, (Milano, Italy, 1987), pp. 287–292.Google Scholar