Grounding deep neural network predictions of human categorization behavior in understandable functional features: the case of face identity

Daube, C., Xu, T., Zhan, J., Webb, A., Ince, R. A.A. , Garrod, O. G.B. and Schyns, P. G. (2021) Grounding deep neural network predictions of human categorization behavior in understandable functional features: the case of face identity. Patterns, 2(10), 100348. (doi: 10.1016/j.patter.2021.100348) (PMCID:PMC8515012)

Abstract

Deep neural networks (DNNs) can resolve real-world categorization tasks with apparent human-level performance. However, true equivalence of behavioral performance between humans and their DNN models requires that their internal mechanisms process equivalent features of the stimulus. To develop such feature equivalence, our methodology leveraged an interpretable and experimentally controlled generative model of the stimuli (realistic three-dimensional textured faces). Humans rated the similarity of randomly generated faces to four familiar identities. We predicted these similarity ratings from the activations of five DNNs trained with different optimization objectives. Using information theoretic redundancy, reverse correlation, and the testing of generalization gradients, we show that DNN predictions of human behavior improve because their shape and texture features overlap with those that subsume human behavior. Thus, we must equate the functional features that subsume the behavioral performances of the brain and its models before comparing where, when, and how these features are processed.

• Item Type: Articles
• Status: Published
• Refereed: Yes
• Glasgow Author(s) Enlighten ID: Garrod, Dr Oliver and Daube, Dr Christoph and Zhan, Dr Jiayu and Xu, Dr Tian and Webb, Dr Andrew and Schyns, Professor Philippe and Ince, Dr Robin
• Authors: Daube, C., Xu, T., Zhan, J., Webb, A., Ince, R. A.A., Garrod, O. G.B., and Schyns, P. G.
• College/School: College of Medical Veterinary and Life Sciences > School of Psychology & Neuroscience
• Journal Name: Patterns
• Publisher: Elsevier (Cell Press)
• ISSN: 2666-3899
• ISSN (Online): 2666-3899
• Published Online: 10 September 2021
• Copyright Holders: Copyright © 2021 The Authors
• First Published: First published in Patterns 2(10): 100348
• Publisher Policy: Reproduced under a Creative Commons License

Related URLs

• University News
• Research Data

Project Code	Award No	Project Name	Principal Investigator	Funder's Name	Funder Ref	Lead Dept
172413		Brain Algorithmics: Reverse Engineering Dynamic Information Processing Networks from MEG time series	Philippe Schyns	Wellcome Trust (WELLCOTR)	107802/Z/15/Z	NP - Centre for Cognitive Neuroimaging (CCNi)
172046		Visual Commonsense for Scene Understanding	Philippe Schyns	Engineering and Physical Sciences Research Council (EPSRC)	EP/N019261/1	NP - Centre for Cognitive Neuroimaging (CCNi)