The Emergence of Organizing Structure in Conceptual Representation

Brenden Lake, Neil D. Lawrence, Joshua B. Tenenbaum

Research output: Contribution to journalArticle

Abstract

Both scientists and children make important structural discoveries, yet their computational underpinnings are not well understood. Structure discovery has previously been formalized as probabilistic inference about the right structural form-where form could be a tree, ring, chain, grid, etc. (Kemp & Tenenbaum, 2008). Although this approach can learn intuitive organizations, including a tree for animals and a ring for the color circle, it assumes a strong inductive bias that considers only these particular forms, and each form is explicitly provided as initial knowledge. Here we introduce a new computational model of how organizing structure can be discovered, utilizing a broad hypothesis space with a preference for sparse connectivity. Given that the inductive bias is more general, the model's initial knowledge shows little qualitative resemblance to some of the discoveries it supports. As a consequence, the model can also learn complex structures for domains that lack intuitive description, as well as predict human property induction judgments without explicit structural forms. By allowing form to emerge from sparsity, our approach clarifies how both the richness and flexibility of human conceptual organization can coexist.

Original languageEnglish (US)
JournalCognitive Science
DOIs
StateAccepted/In press - Jan 1 2018

Fingerprint

Color
Animals

Keywords

  • Bayesian modeling
  • Sparsity
  • Structure discovery
  • Unsupervised learning

ASJC Scopus subject areas

  • Experimental and Cognitive Psychology
  • Cognitive Neuroscience
  • Artificial Intelligence

Cite this

The Emergence of Organizing Structure in Conceptual Representation. / Lake, Brenden; Lawrence, Neil D.; Tenenbaum, Joshua B.

In: Cognitive Science, 01.01.2018.

Research output: Contribution to journalArticle

@article{e1a415cb757a4f518d144349f4a3073a,
title = "The Emergence of Organizing Structure in Conceptual Representation",
abstract = "Both scientists and children make important structural discoveries, yet their computational underpinnings are not well understood. Structure discovery has previously been formalized as probabilistic inference about the right structural form-where form could be a tree, ring, chain, grid, etc. (Kemp & Tenenbaum, 2008). Although this approach can learn intuitive organizations, including a tree for animals and a ring for the color circle, it assumes a strong inductive bias that considers only these particular forms, and each form is explicitly provided as initial knowledge. Here we introduce a new computational model of how organizing structure can be discovered, utilizing a broad hypothesis space with a preference for sparse connectivity. Given that the inductive bias is more general, the model's initial knowledge shows little qualitative resemblance to some of the discoveries it supports. As a consequence, the model can also learn complex structures for domains that lack intuitive description, as well as predict human property induction judgments without explicit structural forms. By allowing form to emerge from sparsity, our approach clarifies how both the richness and flexibility of human conceptual organization can coexist.",
keywords = "Bayesian modeling, Sparsity, Structure discovery, Unsupervised learning",
author = "Brenden Lake and Lawrence, {Neil D.} and Tenenbaum, {Joshua B.}",
year = "2018",
month = "1",
day = "1",
doi = "10.1111/cogs.12580",
language = "English (US)",
journal = "Cognitive Science",
issn = "0364-0213",
publisher = "Wiley-Blackwell",

}

TY - JOUR

T1 - The Emergence of Organizing Structure in Conceptual Representation

AU - Lake, Brenden

AU - Lawrence, Neil D.

AU - Tenenbaum, Joshua B.

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Both scientists and children make important structural discoveries, yet their computational underpinnings are not well understood. Structure discovery has previously been formalized as probabilistic inference about the right structural form-where form could be a tree, ring, chain, grid, etc. (Kemp & Tenenbaum, 2008). Although this approach can learn intuitive organizations, including a tree for animals and a ring for the color circle, it assumes a strong inductive bias that considers only these particular forms, and each form is explicitly provided as initial knowledge. Here we introduce a new computational model of how organizing structure can be discovered, utilizing a broad hypothesis space with a preference for sparse connectivity. Given that the inductive bias is more general, the model's initial knowledge shows little qualitative resemblance to some of the discoveries it supports. As a consequence, the model can also learn complex structures for domains that lack intuitive description, as well as predict human property induction judgments without explicit structural forms. By allowing form to emerge from sparsity, our approach clarifies how both the richness and flexibility of human conceptual organization can coexist.

AB - Both scientists and children make important structural discoveries, yet their computational underpinnings are not well understood. Structure discovery has previously been formalized as probabilistic inference about the right structural form-where form could be a tree, ring, chain, grid, etc. (Kemp & Tenenbaum, 2008). Although this approach can learn intuitive organizations, including a tree for animals and a ring for the color circle, it assumes a strong inductive bias that considers only these particular forms, and each form is explicitly provided as initial knowledge. Here we introduce a new computational model of how organizing structure can be discovered, utilizing a broad hypothesis space with a preference for sparse connectivity. Given that the inductive bias is more general, the model's initial knowledge shows little qualitative resemblance to some of the discoveries it supports. As a consequence, the model can also learn complex structures for domains that lack intuitive description, as well as predict human property induction judgments without explicit structural forms. By allowing form to emerge from sparsity, our approach clarifies how both the richness and flexibility of human conceptual organization can coexist.

KW - Bayesian modeling

KW - Sparsity

KW - Structure discovery

KW - Unsupervised learning

UR - http://www.scopus.com/inward/record.url?scp=85040202665&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85040202665&partnerID=8YFLogxK

U2 - 10.1111/cogs.12580

DO - 10.1111/cogs.12580

M3 - Article

C2 - 29315735

AN - SCOPUS:85040202665

JO - Cognitive Science

JF - Cognitive Science

SN - 0364-0213

ER -