Cross-fertilization between connectionist networks and highly parallel architecturesThe theoretical and practical connections between connectionist schemes such as neural-network computers and traditional symbolic processing architectures involving a high degree of parallelism are explored, reviewing the results of recent investigations. Topics addressed include data flow, data structure, and control flow; conventional pointers; associative addressing; hashing and reduced representations; the problem of binding values to variables; and levels of parallelism. It is concluded that connectionism is more closely related to traditional computer science and technology than is generally admitted; more cooperation between followers of the two approaches is recommended.
Document ID
19900054070
Acquisition Source
Legacy CDMS
Document Type
Conference Paper
Authors
Barnden, John (New Mexico State Univ. Las Cruces, NM, United States)
Srinivas, Kankanahalli (New Mexico State University Las Cruces, United States)