| R. Levinson. Towards domain independent machine intelligence. In Conceptual Graphs for Knowledge Representation. Proc. First International Conference on Conceptual Structures, Quebec, Canada, 1993. Springer-Verlag. |
....term space requirement was the actual memory used to store all codes; for bit vectors, however, the space requirement does not consider memory padding. Still, the improvement that sparse terms offer over bit vectors is remarkable. The first taxonomy was obtained from a chess learning program [95], in which each node is a board position. There are 1,815 nodes (590 meet irreducible elements and 1,425 join irreducibles) and 8,227 links in the transitive reduction. As shown in Table 6.2, sparse terms require one quarter of the space for bit vectors in the top down transitive closure ....
R. Levinson. Towards domain independent machine intelligence. In Conceptual Graphs for Knowledge Representation. Proc. First International Conference on Conceptual Structures, Quebec, Canada, 1993. Springer-Verlag.
....relations, intersections of these may correspond to simpler relations. Further, state variables (corresponding to level 2 in the hierarchy discussed above) can be recognized by starting with all bits in a single variable and gradually splitting into smaller variables by finding counter examples [Lev93]. Such variables correspond very well to unary relations or types in the hidden game definition. For example, these could be pieces or squares in a non malicious encoding of chess. 7 Exploiting analogous relationships In the blind learning framework subsystems will be formed based on relation ....
R. Levinson. Towards domain-independent machine intelligence. In Conceptual Graphs for Knowledge Representation, volume 699 of Lecture Notes in Computer Science, pages 254--273. Springer-Verlag, 1993.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC