Publication:
Uncertainty propagation based MINLP approach for artificial neural network structure reduction

dc.contributor.coauthorŞıldır, Hasan
dc.contributor.coauthorSarrafi, Şahin
dc.contributor.departmentDepartment of Chemical and Biological Engineering
dc.contributor.departmentKUTEM (Koç University Tüpraş Energy Center)
dc.contributor.kuauthorAydın, Erdal
dc.contributor.schoolcollegeinstituteCollege of Engineering
dc.contributor.schoolcollegeinstituteResearch Center
dc.date.accessioned2024-11-09T13:11:28Z
dc.date.issued2022
dc.description.abstractThe performance of artificial neural networks (ANNs) is highly influenced by the selection of input variables and the architecture defined by hyper parameters such as the number of neurons in the hidden layer and connections between network variables. Although there are some black-box and trial and error based studies in the literature to deal with these issues, it is fair to state that a rigorous and systematic method providing global and unique solution is still missing. Accordingly, in this study, a mixed integer nonlinear programming (MINLP) formulation is proposed to detect the best features and connections among the neural network elements while propagating parameter and output uncertainties for regression problems. The objective of the formulation is to minimize the covariance of the estimated parameters while by (i) detecting the ideal number of neurons, (ii) synthesizing the connection configuration between those neurons, inputs and outputs, and (iii) selecting optimum input variables in a multi variable data set to design and ensure identifiable ANN architectures. As a result, suggested approach provides a robust and optimal ANN architecture with tighter prediction bounds obtained from propagation of parameter uncertainty, and higher prediction accuracy compared to the traditional fully connected approach and other benchmarks. Furthermore, such a performance is obtained after elimination of approximately 85% and 90% of the connections, for two case studies respectively, compared to traditional ANN in addition to significant reduction in the input subset.
dc.description.fulltextYES
dc.description.indexedbyWOS
dc.description.indexedbyScopus
dc.description.issue9
dc.description.openaccessYES
dc.description.publisherscopeInternational
dc.description.sponsoredbyTubitakEuTÜBİTAK
dc.description.sponsorshipScientific and Technological Research Council of Turkey (TÜBİTAK)
dc.description.sponsorshipTÜBİTAK 2232 Program
dc.description.sponsorship2232 International Fellowship for Outstanding Researchers Program
dc.description.versionPublisher version
dc.description.volume10
dc.identifier.doi10.3390/pr10091716
dc.identifier.embargoNO
dc.identifier.filenameinventorynoIR04001
dc.identifier.issn2227-9717
dc.identifier.quartileN/A
dc.identifier.scopus2-s2.0-85138734907
dc.identifier.urihttps://hdl.handle.net/20.500.14288/2861
dc.identifier.wos857543300001
dc.keywordsArtificial neural networks
dc.keywordsError propagation
dc.keywordsMixed integer nonlinear programming
dc.keywordsOptimal input selection
dc.keywordsParameter uncertainty
dc.language.isoeng
dc.publisherMultidisciplinary Digital Publishing Institute (MDPI)
dc.relation.grantno118C245
dc.relation.ispartofProcesses
dc.relation.urihttp://cdm21054.contentdm.oclc.org/cdm/ref/collection/IR/id/10885
dc.subjectEngineering
dc.subjectChemical engineering
dc.titleUncertainty propagation based MINLP approach for artificial neural network structure reduction
dc.typeJournal Article
dspace.entity.typePublication
local.contributor.kuauthorAydın, Erdal
local.publication.orgunit1College of Engineering
local.publication.orgunit1Research Center
local.publication.orgunit2KUTEM (Koç University Tüpraş Energy Center)
local.publication.orgunit2Department of Chemical and Biological Engineering
relation.isOrgUnitOfPublicationc747a256-6e0c-4969-b1bf-3b9f2f674289
relation.isOrgUnitOfPublication6ce65247-25c7-415b-a771-a9f0249b3a40
relation.isOrgUnitOfPublication.latestForDiscoveryc747a256-6e0c-4969-b1bf-3b9f2f674289
relation.isParentOrgUnitOfPublication8e756b23-2d4a-4ce8-b1b3-62c794a8c164
relation.isParentOrgUnitOfPublicationd437580f-9309-4ecb-864a-4af58309d287
relation.isParentOrgUnitOfPublication.latestForDiscovery8e756b23-2d4a-4ce8-b1b3-62c794a8c164

Files

Original bundle

Now showing 1 - 1 of 1
Thumbnail Image
Name:
10885.pdf
Size:
2.31 MB
Format:
Adobe Portable Document Format