| Steven Phillips. Constituent similarity and systematicity: the limits of firstorder connectionism. Connection Science, 12(1):45--63, 2000. |
....even at a novel, deeper level of clausal embedding. All network training is unsupervised with respect to error feedback. A novel variant of competitive learning, and an unusual hierarchical architecture are presented. The relationship of this work to issues raised by Marcus (1998) and Phillips (2000) is explored. Keywords: Systematicity, Semantic Features, Language Acquisition, Competitive Learning, Connectionism. Systematicity from Semantic Predictability 2 1. Introduction 0 Within the last decade, simple recurrent networks (SRNs) have attracted considerable notice among connectionists ....
....persist. Some of these misgivings concern the range of syntactic patterns (Marcus, 1998; Rispoli, 1999) that SRN s can actually master. Other doubts involve the degree of systematicity (or word position generalization) that SRNs can attain (see Hadley, 1994a, 1994b, in press; Marcus, 1998; Phillips, 2000). Still other qualms concern the nature of the learning algorithm employed in most SRN experiments, viz. the backpropagation of error. Indeed, Elman and co authors have expressed some misgivings about employing a training algorithm whose biological basis has persistently eluded discovery (Elman ....
[Article contains additional citation context not shown here]
Phillips, S., 2000, Constituent similarity and systematicity: the limits of first-order connectionism. Connection Science, 12: 1-19.
....S##CITEEND##. Halford, G. S. and Hiraki, K. 2000) Proceedings of the Japanese Cognitive Science Society. 1 Infant A not B errors: A case for conceptual dynamics Steven Phillips Information Science Division Electrotechnical Laboratory 1 1 4 Umezono, Tsukuba, 305, Japan stevep etl.go.jp Graeme S. ....
....by looking for generalizations that span experience in a way that necessitates internal structure. For example, generalization over relations between certain perceptual states is not possible with standard function approximation style connectionism without some form of structured representation (Phillips, 2000). In a similar vein, empirical evidence recalled here is not accounted for without supposing some form of internal concept representation. Freeman, Lloyd, and Sinha (1980) provided evidence that occluder functionality (cup used as a container) affects performance. Infants were tested on an ....
Phillips, S. (2000). Constituent similarity and systematicity: The limits of first-order connectionism. Connection Science, 12 (1), 1--19.
No context found.
Steven Phillips. Constituent similarity and systematicity: the limits of firstorder connectionism. Connection Science, 12(1):45--63, 2000.
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