EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Disentangling Abstraction from Statistical Pattern Matching in Human and Machine Learning

Sreejan Kumar iD, Ishita Dasgupta, Nathaniel D Daw, Jonathan D Cohen, Thomas L Griffiths

PLOS Computational Biology · 2023

DOI: 10.1371/journal.pcbi.1011316 ↗

Kurz gefasst

Fachartikel in der Zeitschrift PLOS Computational Biology, 2023. Verfasst von Sreejan Kumar und 4 weiteren Autor:innen. Beteiligte Einrichtung: Neuroscience Institute, Princeton University, Princeton, New Jersey, United States of America. Hauptthemen laut MEDLINE: Machine Learning, Concept Formation. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 59 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: NIMH NIH HHS, NIH, Life Sciences Division, Army Research Office.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The ability to acquire abstract knowledge is a hallmark of human intelligence and is believed by many to be one of the core differences between humans and neural network models. Agents can be endowed with an inductive bias towards abstraction through meta-learning, where they are trained on a distribution of tasks that share some abstract structure that can be learned and applied. However, because neural networks are hard to interpret, it can be difficult to tell whether agents have learned the underlying abstraction, or alternatively statistical patterns that are characteristic of that abstraction. In this work, we compare the performance of humans and agents in a meta-reinforcement learning paradigm in which tasks are generated from abstract rules. We define a novel methodology for building “task metamers” that closely match the statistics of the abstract tasks but use a different underlying generative process, and evaluate performance on both abstract and metamer tasks. We find that humans perform better at abstract tasks than metamer tasks whereas common neural network architectures typically perform worse on the abstract tasks than the matched metamers. This work provides a foundation for characterizing differences between humans and machine learning that can be used in future work towards developing machines with more human-like behavior.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Machine LearningConcept FormationHumansIntelligenceKnowledgeNeural Networks, Computer

Autor:innen und Institutionen

  • Neuroscience Institute, Princeton University, Princeton, New Jersey, United States of America.
  • Google DeepMind, London, United Kingdom.
  • Department of Psychology, Princeton University, Princeton, New Jersey, United States of America.
  • Department of Computer Science, Princeton University, Princeton, New Jersey, United States of America.

ORCID-Profile: Sreejan Kumar

Förderung

  • NIMH NIH HHS
  • NIH
  • Life Sciences Division, Army Research Office
  • Defense Sciences Office, DARPA
  • John Templeton Foundation

Zitieren

APA

Kumar, S., Dasgupta, I., Daw, N. D., Cohen, J. D., & Griffiths, T. L. (2023). Disentangling Abstraction from Statistical Pattern Matching in Human and Machine Learning. PLOS Computational Biology. https://doi.org/10.1371/journal.pcbi.1011316

Vancouver

Kumar S, Dasgupta I, Daw ND, Cohen JD, Griffiths TL. Disentangling Abstraction from Statistical Pattern Matching in Human and Machine Learning. PLOS Computational Biology. 2023. doi:10.1371/journal.pcbi.1011316

BibTeX
@article{Kumar2023,
  author = {Kumar, Sreejan and Dasgupta, Ishita and Daw, Nathaniel D and Cohen, Jonathan D and Griffiths, Thomas L},
  title = {Disentangling Abstraction from Statistical Pattern Matching in Human and Machine Learning},
  journal = {PLOS Computational Biology},
  year = {2023},
  doi = {10.1371/journal.pcbi.1011316}
}
Kostenlos starten →