For deep networks, the whole equals the sum of the parts

Philip J Kellman; Nicholas Baker; Patrick Garrigan; Austin Phillips; Hongjing Lu

doi:10.1017/S0140525X23001541

For deep networks, the whole equals the sum of the parts

Behav Brain Sci. 2023 Dec 6:46:e396. doi: 10.1017/S0140525X23001541.

Authors

Philip J Kellman¹, Nicholas Baker², Patrick Garrigan³, Austin Phillips⁴, Hongjing Lu⁵

Affiliations

¹ Department of Psychology and David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, CA, USA kellman@cognet.ucla.edu; https://kellmanlab.psych.ucla.edu/.
² Department of Psychology, Loyola University of Chicago, Chicago, IL, USA nbaker1@ucla.edu; https://www.luc.edu/psychology/people/staff/facultyandstaff/nicholasbaker/.
³ Department of Psychology, St. Joseph's University, Philadelphia, PA, USA patrick.garrigan@sju.edu; https://sjupsych.org/faculty_pg.php.
⁴ Department of Psychology, University of California, Los Angeles, Los Angeles, CA, USA asphillips@ucla.edu; https://kellmanlab.psych.ucla.edu/.
⁵ Department of Psychology and Department of Statistics, University of California, Los Angeles, Los Angeles, CA, USA hongjing@ucla.edu; https://cvl.psych.ucla.edu/.

PMID: 38054331
DOI: 10.1017/S0140525X23001541

Abstract

Deep convolutional networks exceed humans in sensitivity to local image properties, but unlike biological vision systems, do not discover and encode abstract relations that capture important properties of objects and events in the world. Coupling network architectures with additional machinery for encoding abstract relations will make deep networks better models of human abilities and more versatile and capable artificial devices.

MeSH terms

Deep Learning*
Humans
Neural Networks, Computer*