A highly efficient recurrent neural network architecture for data regression

Simple item page
dc.contributor.authorErgen, Tolgaen_US
dc.contributor.authorCeyani, Emiren_US
dc.coverage.spatialIzmir, Turkeyen_US
dc.date.accessioned2019-02-21T16:05:10Z
dc.date.available2019-02-21T16:05:10Z
dc.date.issued2018en_US
dc.departmentDepartment of Electrical and Electronics Engineeringen_US
dc.descriptionDate of Conference: 2-5 May 2018en_US
dc.description.abstractIn this paper, we study online nonlinear data regression and propose a highly efficient long short term memory (LSTM) network based architecture. Here, we also introduce on-line training algorithms to learn the parameters of the introduced architecture. We first propose an LSTM based architecture for data regression. To diminish the complexity of this architecture, we use an energy efficient operator (ef-operator) instead of the multiplication operation. We then factorize the matrices of the LSTM network to reduce the total number of parameters to be learned. In order to train the parameters of this structure, we introduce online learning methods based on the exponentiated gradient (EG) and stochastic gradient descent (SGD) algorithms. Experimental results demonstrate considerable performance and efficiency improvements provided by the introduced architecture.
dc.description.provenanceMade available in DSpace on 2019-02-21T16:05:10Z (GMT). No. of bitstreams: 1 Bilkent-research-paper.pdf: 222869 bytes, checksum: 842af2b9bd649e7f548593affdbafbb3 (MD5) Previous issue date: 2018en
dc.identifier.doi10.1109/SIU.2018.8404708
dc.identifier.isbn9781538615010
dc.identifier.urihttp://hdl.handle.net/11693/50236
dc.language.isoTurkish
dc.publisherIEEE
dc.relation.isversionofhttps://doi.org/10.1109/SIU.2018.8404708
dc.source.title2018 26th Signal Processing and Communications Applications Conference (SIU)en_US
dc.subjectEf-operatoren_US
dc.subjectExponentiated gradienten_US
dc.subjectGradient descenten_US
dc.subjectLong short term memory networken_US
dc.subjectMatrix factorizationen_US
dc.titleA highly efficient recurrent neural network architecture for data regressionen_US
dc.title.alternativeVeri bağlanımı için yüksek verimli yinelemeli sinir ağı yapısıen_US
dc.typeConference Paperen_US

Files

Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
A_highly_efficient_recurrent_neural_network_architecture_for_data_regression.pdf
Size:
499.74 KB
Format:
Adobe Portable Document Format
Description:
Full printable version
Download

Collections

Department of Electrical and Electronics Engineering