Superpositional connectionism: A reply to Marinov

Research output: Contribution to journalLetterResearchpeer-review

Abstract

Marinov's critique I argue, is vitiated by its failure to recognize the distinctive role of superposition within the distributed connectionist paradigm. The use of so-called 'subsymbolic' distributed encodings alone is not, I agree, enough to justify treating distributed connectionism as a distinctive approach. It has always been clear that microfeatural decomposition is both possible and actual within the confines of recognizably classical approaches. When such approaches also involve statistically-driven learning algorithms - as in the case of ID3 - the fundamental differences become even harder to spot. To see them, it is necessary to consider not just the nature of an acquired input-output function but the nature of the representational scheme underlying it. Differences between such schemes make themselves best felt outside the domain of immediate problem solving. It is in the more extended contexts of performance DURING learning and cognitive change as a result of SUBSEQUENT training on new tasks (or simultaneous training on several tasks) that the effects of superpositional storage techniques come to the fore. I conclude that subsymbols, distribution and statistically driven learning alone are indeed not of the essence. But connectionism is not just about subsymbols and distribution. It is about the generation of whole subsymbol SYSTEMS in which multiple distributed representations are created and superposed.

LanguageEnglish
Pages271-281
Number of pages11
JournalMinds and Machines
Volume3
Issue number3
DOIs
Publication statusPublished - Aug 1993
Externally publishedYes

Fingerprint

Learning algorithms
Decomposition
Connectionism
Essence
Problem Solving
Paradigm
Distributed Representation
Superposition
Encoding
Connectionist
Fundamental

Keywords

  • Connectionism
  • distribution
  • subsymbol
  • symbol

Cite this

@article{913f3d8c4a944457b057c5a80413cdd6,
title = "Superpositional connectionism: A reply to Marinov",
abstract = "Marinov's critique I argue, is vitiated by its failure to recognize the distinctive role of superposition within the distributed connectionist paradigm. The use of so-called 'subsymbolic' distributed encodings alone is not, I agree, enough to justify treating distributed connectionism as a distinctive approach. It has always been clear that microfeatural decomposition is both possible and actual within the confines of recognizably classical approaches. When such approaches also involve statistically-driven learning algorithms - as in the case of ID3 - the fundamental differences become even harder to spot. To see them, it is necessary to consider not just the nature of an acquired input-output function but the nature of the representational scheme underlying it. Differences between such schemes make themselves best felt outside the domain of immediate problem solving. It is in the more extended contexts of performance DURING learning and cognitive change as a result of SUBSEQUENT training on new tasks (or simultaneous training on several tasks) that the effects of superpositional storage techniques come to the fore. I conclude that subsymbols, distribution and statistically driven learning alone are indeed not of the essence. But connectionism is not just about subsymbols and distribution. It is about the generation of whole subsymbol SYSTEMS in which multiple distributed representations are created and superposed.",
keywords = "Connectionism, distribution, subsymbol, symbol",
author = "Andy Clark",
year = "1993",
month = "8",
doi = "10.1007/BF00976282",
language = "English",
volume = "3",
pages = "271--281",
journal = "Minds and Machines",
issn = "0924-6495",
publisher = "Springer, Springer Nature",
number = "3",

}

Superpositional connectionism : A reply to Marinov. / Clark, Andy.

In: Minds and Machines, Vol. 3, No. 3, 08.1993, p. 271-281.

Research output: Contribution to journalLetterResearchpeer-review

TY - JOUR

T1 - Superpositional connectionism

T2 - Minds and Machines

AU - Clark, Andy

PY - 1993/8

Y1 - 1993/8

N2 - Marinov's critique I argue, is vitiated by its failure to recognize the distinctive role of superposition within the distributed connectionist paradigm. The use of so-called 'subsymbolic' distributed encodings alone is not, I agree, enough to justify treating distributed connectionism as a distinctive approach. It has always been clear that microfeatural decomposition is both possible and actual within the confines of recognizably classical approaches. When such approaches also involve statistically-driven learning algorithms - as in the case of ID3 - the fundamental differences become even harder to spot. To see them, it is necessary to consider not just the nature of an acquired input-output function but the nature of the representational scheme underlying it. Differences between such schemes make themselves best felt outside the domain of immediate problem solving. It is in the more extended contexts of performance DURING learning and cognitive change as a result of SUBSEQUENT training on new tasks (or simultaneous training on several tasks) that the effects of superpositional storage techniques come to the fore. I conclude that subsymbols, distribution and statistically driven learning alone are indeed not of the essence. But connectionism is not just about subsymbols and distribution. It is about the generation of whole subsymbol SYSTEMS in which multiple distributed representations are created and superposed.

AB - Marinov's critique I argue, is vitiated by its failure to recognize the distinctive role of superposition within the distributed connectionist paradigm. The use of so-called 'subsymbolic' distributed encodings alone is not, I agree, enough to justify treating distributed connectionism as a distinctive approach. It has always been clear that microfeatural decomposition is both possible and actual within the confines of recognizably classical approaches. When such approaches also involve statistically-driven learning algorithms - as in the case of ID3 - the fundamental differences become even harder to spot. To see them, it is necessary to consider not just the nature of an acquired input-output function but the nature of the representational scheme underlying it. Differences between such schemes make themselves best felt outside the domain of immediate problem solving. It is in the more extended contexts of performance DURING learning and cognitive change as a result of SUBSEQUENT training on new tasks (or simultaneous training on several tasks) that the effects of superpositional storage techniques come to the fore. I conclude that subsymbols, distribution and statistically driven learning alone are indeed not of the essence. But connectionism is not just about subsymbols and distribution. It is about the generation of whole subsymbol SYSTEMS in which multiple distributed representations are created and superposed.

KW - Connectionism

KW - distribution

KW - subsymbol

KW - symbol

UR - http://www.scopus.com/inward/record.url?scp=34250074959&partnerID=8YFLogxK

U2 - 10.1007/BF00976282

DO - 10.1007/BF00976282

M3 - Letter

VL - 3

SP - 271

EP - 281

JO - Minds and Machines

JF - Minds and Machines

SN - 0924-6495

IS - 3

ER -