No Cover Image

Conference Paper/Proceeding/Abstract 395 views

Weight Perturbations for Simulating Virtual Lesions in a Convolutional Neural Network

Joe MacInnes Orcid Logo, Natalia Zhozhikashvili, Matteo Feurra

Lecture Notes in Computer Science, Volume: 14976, Pages: 221 - 234

Swansea University Author: Joe MacInnes Orcid Logo

Full text not available from this repository: check for access using links below.

Check full text

DOI (Published version): 10.1007/978-3-031-67285-9_16

Abstract

Convolutional Neural Networks (CNNs) match human performance in many visual tasks like the classification of images, however they may not simulate the underlying biological processes. We implemented a CNN to try replicate results from an object inversion experiment with Transcranial Magnetic Stimula...

Full description

Published in: Lecture Notes in Computer Science
ISBN: 9783031672842 9783031672859
ISSN: 0302-9743 1611-3349
Published: Cham Springer Nature Switzerland 2024
Online Access: Check full text

URI: https://cronfa.swan.ac.uk/Record/cronfa69412
first_indexed 2025-05-02T14:29:00Z
last_indexed 2025-06-27T09:24:51Z
id cronfa69412
recordtype SURis
fullrecord <?xml version="1.0"?><rfc1807><datestamp>2025-06-26T15:37:54.8084513</datestamp><bib-version>v2</bib-version><id>69412</id><entry>2025-05-02</entry><title>Weight Perturbations for Simulating Virtual Lesions in a Convolutional Neural Network</title><swanseaauthors><author><sid>06dcb003ec50192bafde2c77bef4fd5c</sid><ORCID>0000-0002-5134-1601</ORCID><firstname>Joe</firstname><surname>MacInnes</surname><name>Joe MacInnes</name><active>true</active><ethesisStudent>false</ethesisStudent></author></swanseaauthors><date>2025-05-02</date><deptcode>MACS</deptcode><abstract>Convolutional Neural Networks (CNNs) match human performance in many visual tasks like the classification of images, however they may not simulate the underlying biological processes. We implemented a CNN to try replicate results from an object inversion experiment with Transcranial Magnetic Stimulation (TMS). After training on upright faces, the CNN model went through three stages of testing: checking (1) for overall accuracy, (2) for the presence of the Face Inversion Effect (FIE) and (3) for an FIE reduction after weight perturbations. Results of the model were compared with human performance in an analogous experiment, where disruption of the extrastriate cortex (the Occipital Face Area (OFA) and the control Occipital Place Area (OPA)) was performed using TMS. The resulting model (1) showed a level of accuracy similar to humans, but (2) did not show the FIE, but rather showed a general object inverted effect. Disruption with TMS (3) led to a reduction in the FIE, however disruption of model layers only led to reduction of the general object inverted effect. Thus, CNNs were observed to successfully simulate some results of objects recognition in general, but are unable to simulate the specific mechanisms of modularity and face processing. CNNs are certainly a useful metaphor for human visual processing, but it&#x2019;s important to understand the limits of that metaphor if they are to be used as models in medicine and neuroscience.</abstract><type>Conference Paper/Proceeding/Abstract</type><journal>Lecture Notes in Computer Science</journal><volume>14976</volume><journalNumber/><paginationStart>221</paginationStart><paginationEnd>234</paginationEnd><publisher>Springer Nature Switzerland</publisher><placeOfPublication>Cham</placeOfPublication><isbnPrint>9783031672842</isbnPrint><isbnElectronic>9783031672859</isbnElectronic><issnPrint>0302-9743</issnPrint><issnElectronic>1611-3349</issnElectronic><keywords/><publishedDay>15</publishedDay><publishedMonth>8</publishedMonth><publishedYear>2024</publishedYear><publishedDate>2024-08-15</publishedDate><doi>10.1007/978-3-031-67285-9_16</doi><url/><notes/><college>COLLEGE NANME</college><department>Mathematics and Computer Science School</department><CollegeCode>COLLEGE CODE</CollegeCode><DepartmentCode>MACS</DepartmentCode><institution>Swansea University</institution><apcterm/><funders/><projectreference/><lastEdited>2025-06-26T15:37:54.8084513</lastEdited><Created>2025-05-02T15:26:39.6963385</Created><path><level id="1">Faculty of Science and Engineering</level><level id="2">School of Mathematics and Computer Science - Computer Science</level></path><authors><author><firstname>Joe</firstname><surname>MacInnes</surname><orcid>0000-0002-5134-1601</orcid><order>1</order></author><author><firstname>Natalia</firstname><surname>Zhozhikashvili</surname><order>2</order></author><author><firstname>Matteo</firstname><surname>Feurra</surname><order>3</order></author></authors><documents/><OutputDurs/></rfc1807>
spelling 2025-06-26T15:37:54.8084513 v2 69412 2025-05-02 Weight Perturbations for Simulating Virtual Lesions in a Convolutional Neural Network 06dcb003ec50192bafde2c77bef4fd5c 0000-0002-5134-1601 Joe MacInnes Joe MacInnes true false 2025-05-02 MACS Convolutional Neural Networks (CNNs) match human performance in many visual tasks like the classification of images, however they may not simulate the underlying biological processes. We implemented a CNN to try replicate results from an object inversion experiment with Transcranial Magnetic Stimulation (TMS). After training on upright faces, the CNN model went through three stages of testing: checking (1) for overall accuracy, (2) for the presence of the Face Inversion Effect (FIE) and (3) for an FIE reduction after weight perturbations. Results of the model were compared with human performance in an analogous experiment, where disruption of the extrastriate cortex (the Occipital Face Area (OFA) and the control Occipital Place Area (OPA)) was performed using TMS. The resulting model (1) showed a level of accuracy similar to humans, but (2) did not show the FIE, but rather showed a general object inverted effect. Disruption with TMS (3) led to a reduction in the FIE, however disruption of model layers only led to reduction of the general object inverted effect. Thus, CNNs were observed to successfully simulate some results of objects recognition in general, but are unable to simulate the specific mechanisms of modularity and face processing. CNNs are certainly a useful metaphor for human visual processing, but it’s important to understand the limits of that metaphor if they are to be used as models in medicine and neuroscience. Conference Paper/Proceeding/Abstract Lecture Notes in Computer Science 14976 221 234 Springer Nature Switzerland Cham 9783031672842 9783031672859 0302-9743 1611-3349 15 8 2024 2024-08-15 10.1007/978-3-031-67285-9_16 COLLEGE NANME Mathematics and Computer Science School COLLEGE CODE MACS Swansea University 2025-06-26T15:37:54.8084513 2025-05-02T15:26:39.6963385 Faculty of Science and Engineering School of Mathematics and Computer Science - Computer Science Joe MacInnes 0000-0002-5134-1601 1 Natalia Zhozhikashvili 2 Matteo Feurra 3
title Weight Perturbations for Simulating Virtual Lesions in a Convolutional Neural Network
spellingShingle Weight Perturbations for Simulating Virtual Lesions in a Convolutional Neural Network
Joe MacInnes
title_short Weight Perturbations for Simulating Virtual Lesions in a Convolutional Neural Network
title_full Weight Perturbations for Simulating Virtual Lesions in a Convolutional Neural Network
title_fullStr Weight Perturbations for Simulating Virtual Lesions in a Convolutional Neural Network
title_full_unstemmed Weight Perturbations for Simulating Virtual Lesions in a Convolutional Neural Network
title_sort Weight Perturbations for Simulating Virtual Lesions in a Convolutional Neural Network
author_id_str_mv 06dcb003ec50192bafde2c77bef4fd5c
author_id_fullname_str_mv 06dcb003ec50192bafde2c77bef4fd5c_***_Joe MacInnes
author Joe MacInnes
author2 Joe MacInnes
Natalia Zhozhikashvili
Matteo Feurra
format Conference Paper/Proceeding/Abstract
container_title Lecture Notes in Computer Science
container_volume 14976
container_start_page 221
publishDate 2024
institution Swansea University
isbn 9783031672842
9783031672859
issn 0302-9743
1611-3349
doi_str_mv 10.1007/978-3-031-67285-9_16
publisher Springer Nature Switzerland
college_str Faculty of Science and Engineering
hierarchytype
hierarchy_top_id facultyofscienceandengineering
hierarchy_top_title Faculty of Science and Engineering
hierarchy_parent_id facultyofscienceandengineering
hierarchy_parent_title Faculty of Science and Engineering
department_str School of Mathematics and Computer Science - Computer Science{{{_:::_}}}Faculty of Science and Engineering{{{_:::_}}}School of Mathematics and Computer Science - Computer Science
document_store_str 0
active_str 0
description Convolutional Neural Networks (CNNs) match human performance in many visual tasks like the classification of images, however they may not simulate the underlying biological processes. We implemented a CNN to try replicate results from an object inversion experiment with Transcranial Magnetic Stimulation (TMS). After training on upright faces, the CNN model went through three stages of testing: checking (1) for overall accuracy, (2) for the presence of the Face Inversion Effect (FIE) and (3) for an FIE reduction after weight perturbations. Results of the model were compared with human performance in an analogous experiment, where disruption of the extrastriate cortex (the Occipital Face Area (OFA) and the control Occipital Place Area (OPA)) was performed using TMS. The resulting model (1) showed a level of accuracy similar to humans, but (2) did not show the FIE, but rather showed a general object inverted effect. Disruption with TMS (3) led to a reduction in the FIE, however disruption of model layers only led to reduction of the general object inverted effect. Thus, CNNs were observed to successfully simulate some results of objects recognition in general, but are unable to simulate the specific mechanisms of modularity and face processing. CNNs are certainly a useful metaphor for human visual processing, but it’s important to understand the limits of that metaphor if they are to be used as models in medicine and neuroscience.
published_date 2024-08-15T05:30:00Z
_version_ 1856896185278660608
score 11.096068

Similar Items