RECURSIVE DISTRIBUTED REPRESENTATIONS

被引:403
作者
POLLACK, JB [1 ]
机构
[1] OHIO STATE UNIV,DEPT COMP & INFORMAT SCI,COLUMBUS,OH 43210
关键词
D O I
10.1016/0004-3702(90)90005-K
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
A longstanding difficulty for connectionist modeling has been how to represent variable-sized recursive data structures, such as trees and lists, in fixed-width patterns. This paper presents a connectionist architecture which automatically develops compact distributed representations for such compositional structures, as well as efficient accessing mechanisms for them. Patterns which stand for the internal nodes of fixed-valence trees are devised through the recursive use of backpropagation on three-layer auto-associative encoder networks. The resulting representations are novel, in that they combine apparently immiscible aspects of features, pointers, and symbol structures. They form a bridge between the data structures necessary for high-level cognitive tasks and the associative, pattern recognition machinery provided by neural networks. © 1990.
引用
收藏
页码:77 / 105
页数:29
相关论文
共 41 条
[11]  
Hinton G.E., 1986, P 8 ANN C COGN SCI S, V1, P12
[12]  
HINTON GE, 1988, 10TH P ANN C COGN SC, P48
[13]  
HINTON GE, 1984, CMUCS84119 CARN MELL
[14]  
Holland J.H., 1986, INDUCTION PROCESSES
[15]  
HORNIK K, INPRESS NEURAL NETWO
[16]  
JORDAN MI, 1986, ICS8608 U CAL I COGN
[17]  
KAWAMOTO A, 1985, THESIS BROWN U PROVI
[18]  
LAPEDES AS, 1988, LAUR88418 TECH REPT
[19]  
LEHNERT WG, 1987, 6TH P NAT C ART INT, P301
[20]  
Lippmann R. P., 1987, IEEE ASSP Magazine, V4, P4, DOI 10.1145/44571.44572