Deep neural networks are not a single hypothesis but a language for expressing computational hypotheses

Tal Golan; John Mark Taylor; Heiko Herbert Schütt; Benjamin Peters; Rowan P. Sommers; Katja Seeliger; Adrien Doerig; Paul Linton; Talia Konkle; Marcel van Gerven; Konrad Kording; Blake Richards; Tim C. Kietzmann; Grace W. Lindsay; Nikolaus Kriegeskorte

doi:10.1017/S0140525X23001553

Deep neural networks are not a single hypothesis but a language for expressing computational hypotheses

Tal Golan, John Mark Taylor, Heiko Herbert Schütt, Benjamin Peters, Rowan P. Sommers, Katja Seeliger, Adrien Doerig, Paul Linton, Talia Konkle, Marcel van Gerven, Konrad Kording, Blake Richards, Tim C. Kietzmann, Grace W. Lindsay, Nikolaus Kriegeskorte

Research output: Contribution to journal › Comment/debate

Abstract

An ideal vision model accounts for behavior and neurophysiology in both naturalistic conditions and designed lab experiments. Unlike psychological theories, artificial neural networks (ANNs) actually perform visual tasks and generate testable predictions for arbitrary inputs. These advantages enable ANNs to engage the entire spectrum of the evidence. Failures of particular models drive progress in a vibrant ANN research program of human vision.

Original language	English
Article number	e392
Pages (from-to)	e392
Journal	The Behavioral and brain sciences
Volume	46
DOIs	https://doi.org/10.1017/S0140525X23001553
State	Published - 6 Dec 2023

ASJC Scopus subject areas

Neuropsychology and Physiological Psychology
Physiology
Behavioral Neuroscience

Access to Document

10.1017/S0140525X23001553

Cite this

Golan, T., Taylor, J. M., Schütt, H. H., Peters, B., Sommers, R. P., Seeliger, K., Doerig, A., Linton, P., Konkle, T., van Gerven, M., Kording, K., Richards, B., Kietzmann, T. C., Lindsay, G. W., & Kriegeskorte, N. (2023). Deep neural networks are not a single hypothesis but a language for expressing computational hypotheses. The Behavioral and brain sciences, 46, e392. Article e392. https://doi.org/10.1017/S0140525X23001553

@article{4dbe6c34eee8418a88a801ee3e88e6c4,

title = "Deep neural networks are not a single hypothesis but a language for expressing computational hypotheses",

abstract = "An ideal vision model accounts for behavior and neurophysiology in both naturalistic conditions and designed lab experiments. Unlike psychological theories, artificial neural networks (ANNs) actually perform visual tasks and generate testable predictions for arbitrary inputs. These advantages enable ANNs to engage the entire spectrum of the evidence. Failures of particular models drive progress in a vibrant ANN research program of human vision.",

author = "Tal Golan and Taylor, {John Mark} and Sch{\"u}tt, {Heiko Herbert} and Benjamin Peters and Sommers, {Rowan P.} and Katja Seeliger and Adrien Doerig and Paul Linton and Talia Konkle and {van Gerven}, Marcel and Konrad Kording and Blake Richards and Kietzmann, {Tim C.} and Lindsay, {Grace W.} and Nikolaus Kriegeskorte",

note = "Publisher Copyright: Copyright {\textcopyright} The Author(s), 2023.",

year = "2023",

month = dec,

day = "6",

doi = "10.1017/S0140525X23001553",

language = "English",

volume = "46",

pages = "e392",

journal = "The Behavioral and brain sciences",

issn = "0140-525X",

publisher = "Cambridge University Press",

}

Golan, T, Taylor, JM, Schütt, HH, Peters, B, Sommers, RP, Seeliger, K, Doerig, A, Linton, P, Konkle, T, van Gerven, M, Kording, K, Richards, B, Kietzmann, TC, Lindsay, GW & Kriegeskorte, N 2023, 'Deep neural networks are not a single hypothesis but a language for expressing computational hypotheses', The Behavioral and brain sciences, vol. 46, e392, pp. e392. https://doi.org/10.1017/S0140525X23001553

TY - JOUR

T1 - Deep neural networks are not a single hypothesis but a language for expressing computational hypotheses

AU - Golan, Tal

AU - Taylor, John Mark

AU - Schütt, Heiko Herbert

AU - Peters, Benjamin

AU - Sommers, Rowan P.

AU - Seeliger, Katja

AU - Doerig, Adrien

AU - Linton, Paul

AU - Konkle, Talia

AU - van Gerven, Marcel

AU - Kording, Konrad

AU - Richards, Blake

AU - Kietzmann, Tim C.

AU - Lindsay, Grace W.

AU - Kriegeskorte, Nikolaus

PY - 2023/12/6

Y1 - 2023/12/6

N2 - An ideal vision model accounts for behavior and neurophysiology in both naturalistic conditions and designed lab experiments. Unlike psychological theories, artificial neural networks (ANNs) actually perform visual tasks and generate testable predictions for arbitrary inputs. These advantages enable ANNs to engage the entire spectrum of the evidence. Failures of particular models drive progress in a vibrant ANN research program of human vision.

AB - An ideal vision model accounts for behavior and neurophysiology in both naturalistic conditions and designed lab experiments. Unlike psychological theories, artificial neural networks (ANNs) actually perform visual tasks and generate testable predictions for arbitrary inputs. These advantages enable ANNs to engage the entire spectrum of the evidence. Failures of particular models drive progress in a vibrant ANN research program of human vision.

UR - http://www.scopus.com/inward/record.url?scp=85179022645&partnerID=8YFLogxK

U2 - 10.1017/S0140525X23001553

DO - 10.1017/S0140525X23001553

M3 - Comment/debate

C2 - 38054329

AN - SCOPUS:85179022645

SN - 0140-525X

VL - 46

SP - e392

JO - The Behavioral and brain sciences

JF - The Behavioral and brain sciences

M1 - e392

ER -

Deep neural networks are not a single hypothesis but a language for expressing computational hypotheses

Abstract

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this