| O. Melnik. Representation of Information in Neural Networks. PhD thesis, Brandeis University, 2000. |
....any number of dimensions. 90 8.2 Snapshot of a Growing Lexical Model. Labels and dots denote existing nodes. Reproduced, with permission, from [31] 92 xviii Introduction 1. 1 An adequate substrate for symbolic computation Melnik [64] provides the following key insight about neural networks and computational paradigms in general: Even though neural networks are an example of a Turing equivalent computational system [68] in the real world, this theoretical equivalence is of only minor signi cance, as di erent ....
O. Melnik. Representation of Information in Neural Networks. PhD thesis, Brandeis University, 2000.
.... interest, they do not address the degree to which a particular computational paradigm (connectionism) is suited to a particular realworld task (language) They are therefore not of much use in arguing for or against the merits of connectionism as a model of any particular domain of interest (Melnik 2000), any more than knowing about Turing equivalence will help you in choosing between a Macintosh and a Pentium based PC. The third approach, which some of its proponents have described as Representations without Rules (Horgan and Tienson 1989) is the one that we wish to take here. This approach ....
Melnik, O. (2000). Representation of Information in Neural Networks. Ph. D. thesis, Brandeis University.
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