Do Additional Features Help or Hurt Category Learning?

The Curse of Dimensionality in Human Learners

Wai Keen Vong, Andrew T Hendrickson, Danielle J Navarro, Amy Perfors

Research output: Contribution to journalArticleScientificpeer-review

Abstract

The curse of dimensionality, which has been widely studied in statistics and machine learning, occurs when additional features cause the size of the feature space to grow so quickly that learning classification rules becomes increasingly difficult. How do people overcome the curse of dimensionality when acquiring real-world categories that have many different features? Here we investigate the possibility that the structure of categories can help. We show that when categories follow a family resemblance structure, people are unaffected by the presence of additional features in learning. However, when categories are based on a single feature, they fall prey to the curse, and having additional irrelevant features hurts performance. We compare and contrast these results to three different computational models to show that a model with limited computational capacity best captures human performance across almost all of the conditions in both experiments.

Original languageEnglish
Article number12724
Pages (from-to)e12724
Number of pages25
JournalCognitive Science
Volume43
Issue number3
DOIs
Publication statusPublished - Mar 2019

Fingerprint

Learning systems
Statistics
Experiments

Keywords

  • CATEGORIZATION
  • CLASSIFICATION
  • Category learning
  • Curse of dimensionality
  • IDENTIFICATION
  • KNOWLEDGE
  • MODELS
  • Supervised learning

Cite this

Vong, Wai Keen ; Hendrickson, Andrew T ; Navarro, Danielle J ; Perfors, Amy. / Do Additional Features Help or Hurt Category Learning? The Curse of Dimensionality in Human Learners. In: Cognitive Science. 2019 ; Vol. 43, No. 3. pp. e12724.
@article{793522e255d24da9b8245fa7e7db6478,
title = "Do Additional Features Help or Hurt Category Learning?: The Curse of Dimensionality in Human Learners",
abstract = "The curse of dimensionality, which has been widely studied in statistics and machine learning, occurs when additional features cause the size of the feature space to grow so quickly that learning classification rules becomes increasingly difficult. How do people overcome the curse of dimensionality when acquiring real-world categories that have many different features? Here we investigate the possibility that the structure of categories can help. We show that when categories follow a family resemblance structure, people are unaffected by the presence of additional features in learning. However, when categories are based on a single feature, they fall prey to the curse, and having additional irrelevant features hurts performance. We compare and contrast these results to three different computational models to show that a model with limited computational capacity best captures human performance across almost all of the conditions in both experiments.",
keywords = "CATEGORIZATION, CLASSIFICATION, Category learning, Curse of dimensionality, IDENTIFICATION, KNOWLEDGE, MODELS, Supervised learning",
author = "Vong, {Wai Keen} and Hendrickson, {Andrew T} and Navarro, {Danielle J} and Amy Perfors",
note = "{\circledC} 2019 Cognitive Science Society, Inc.",
year = "2019",
month = "3",
doi = "10.1111/cogs.12724",
language = "English",
volume = "43",
pages = "e12724",
journal = "Cognitive Science",
issn = "0364-0213",
publisher = "Wiley",
number = "3",

}

Do Additional Features Help or Hurt Category Learning? The Curse of Dimensionality in Human Learners. / Vong, Wai Keen; Hendrickson, Andrew T; Navarro, Danielle J; Perfors, Amy.

In: Cognitive Science, Vol. 43, No. 3, 12724, 03.2019, p. e12724.

Research output: Contribution to journalArticleScientificpeer-review

TY - JOUR

T1 - Do Additional Features Help or Hurt Category Learning?

T2 - The Curse of Dimensionality in Human Learners

AU - Vong, Wai Keen

AU - Hendrickson, Andrew T

AU - Navarro, Danielle J

AU - Perfors, Amy

N1 - © 2019 Cognitive Science Society, Inc.

PY - 2019/3

Y1 - 2019/3

N2 - The curse of dimensionality, which has been widely studied in statistics and machine learning, occurs when additional features cause the size of the feature space to grow so quickly that learning classification rules becomes increasingly difficult. How do people overcome the curse of dimensionality when acquiring real-world categories that have many different features? Here we investigate the possibility that the structure of categories can help. We show that when categories follow a family resemblance structure, people are unaffected by the presence of additional features in learning. However, when categories are based on a single feature, they fall prey to the curse, and having additional irrelevant features hurts performance. We compare and contrast these results to three different computational models to show that a model with limited computational capacity best captures human performance across almost all of the conditions in both experiments.

AB - The curse of dimensionality, which has been widely studied in statistics and machine learning, occurs when additional features cause the size of the feature space to grow so quickly that learning classification rules becomes increasingly difficult. How do people overcome the curse of dimensionality when acquiring real-world categories that have many different features? Here we investigate the possibility that the structure of categories can help. We show that when categories follow a family resemblance structure, people are unaffected by the presence of additional features in learning. However, when categories are based on a single feature, they fall prey to the curse, and having additional irrelevant features hurts performance. We compare and contrast these results to three different computational models to show that a model with limited computational capacity best captures human performance across almost all of the conditions in both experiments.

KW - CATEGORIZATION

KW - CLASSIFICATION

KW - Category learning

KW - Curse of dimensionality

KW - IDENTIFICATION

KW - KNOWLEDGE

KW - MODELS

KW - Supervised learning

U2 - 10.1111/cogs.12724

DO - 10.1111/cogs.12724

M3 - Article

VL - 43

SP - e12724

JO - Cognitive Science

JF - Cognitive Science

SN - 0364-0213

IS - 3

M1 - 12724

ER -