TY - CONF
T1 - Massively Parallel Knowledge Representation
AU - Geller, James
N1 - Funding Information:
Class hierarchies form the backbone of all frame systems, most semantic network theories, and all object-oriented languages, databases and systems. Their computational investigation goes back to Quillian [1] and forms a main topic of research in AI, Knowledge Representation [2, 3], Databases [4], and Object-Oriented systems [5]. Surprisingly, AI researchers had not given much thought to efficient implementations of large class hierarchies until Schubert [7] introduced a coding that permitted the fast verification or denial of a subclass relation between two given classes. The fast, C~reflexive" [8] responses of human subjects to questions involving IS-A hierarchies led some researchers such as Schubert and Shastri [9, 10] to exploit class hierarchies using special purpose reasoners. AI researchers have, in general, not extended this approach to other transitive relations. Humans can respond to questions such as, "What is older, the pyramids or the US constitution"? with equal "reflexive" speed. The same applies to relations such as taller, faster, heavier, and so on. In the (relational) database literature, transitive closure operations for other relations have drawn considerable attention [11, 19.]. To achieve fast response times for special purpose IS-A reasoners, a few AI researchers have made use of massive parallelism. Shastri [10] uses a connectionist massively parallel method, while Evett, Hendler *This work was partially supported by the National Science Foundation, Grant #IRI-9204655. This work was conducted using the computationalr esources of the Northeast Parallel Architectures Center (NP.kC) at Syracuse University, which funded by and operates under contract to DARP.ka nd the Air Force Systems Corm’hand, RomeA ir Development Center (RADC),G rifllss Air Force Base, NY, under contract# F306002-88-C-0031.
Funding Information:
This work was partially supported by the National Science Foundation, Grant #IRI-9204655. This work was conducted using the computational resources of the Northeast Parallel Architectures Center (NPAC) at Syracuse University, which is funded by and operates under contract to DARPA and the Air Force Systems Command, Rome Air Development Center (RADC), Griffiss Air Force Base, NY, under contract# F306002-88-C-0031.
Publisher Copyright:
© 1993 AI Access Foundation. All rights reserved.
PY - 1993
Y1 - 1993
N2 - We piesent a massively parallel representation of transitive relations, emphasizing the subclass relation, which extends our previous linear tree representation of class hierarchies. This representation makes use of a grid of processors and distributes information about one class over the processors of one column of the grid. As such, it can deal with a subset of directed acyclic graphs. We prove that a "node insertion" can be performed efficiently in our representation.
AB - We piesent a massively parallel representation of transitive relations, emphasizing the subclass relation, which extends our previous linear tree representation of class hierarchies. This representation makes use of a grid of processors and distributes information about one class over the processors of one column of the grid. As such, it can deal with a subset of directed acyclic graphs. We prove that a "node insertion" can be performed efficiently in our representation.
UR - http://www.scopus.com/inward/record.url?scp=77957042425&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=77957042425&partnerID=8YFLogxK
M3 - Paper
AN - SCOPUS:77957042425
SP - 90
EP - 97
T2 - 1993 AAAI Spring Symposium
Y2 - 23 March 1993 through 25 March 1993
ER -