000889208 001__ 889208
000889208 005__ 20210127115307.0
000889208 0247_ $$2doi$$a10.1007/978-3-030-64580-9_7
000889208 0247_ $$2Handle$$a2128/26777
000889208 037__ $$aFZJ-2021-00117
000889208 041__ $$aEnglish
000889208 1001_ $$0P:(DE-Juel1)161462$$aYegenoglu, Alper$$b0$$eCorresponding author$$ufzj
000889208 1112_ $$aThe Sixth International Conference on Machine Learning, Optimization, and Data Science$$cSiena$$d2020-07-19 - 2020-07-22$$gLOD2020$$wItaly
000889208 245__ $$aEnsemble Kalman Filter Optimizing Deep Neural Networks: An Alternative Approach to Non-performing Gradient Descent
000889208 250__ $$a5th ed.
000889208 260__ $$aCham$$bSpringer$$c2020
000889208 29510 $$aMachine Learning, Optimization, and Data Science
000889208 300__ $$a78-92
000889208 3367_ $$2ORCID$$aCONFERENCE_PAPER
000889208 3367_ $$033$$2EndNote$$aConference Paper
000889208 3367_ $$2BibTeX$$aINPROCEEDINGS
000889208 3367_ $$2DRIVER$$aconferenceObject
000889208 3367_ $$2DataCite$$aOutput Types/Conference Paper
000889208 3367_ $$0PUB:(DE-HGF)8$$2PUB:(DE-HGF)$$aContribution to a conference proceedings$$bcontrib$$mcontrib$$s1610561372_10824
000889208 3367_ $$0PUB:(DE-HGF)7$$2PUB:(DE-HGF)$$aContribution to a book$$mcontb
000889208 4900_ $$aLecture Notes in Computer Science$$v12566
000889208 520__ $$aThe successful training of deep neural networks is dependent on initialization schemes and choice of activation functions. Non-optimally chosen parameter settings lead to the known problem of exploding or vanishing gradients. This issue occurs when gradient descent and backpropagation are applied. For this setting the Ensemble Kalman Filter (EnKF) can be used as an alternative optimizer when training neural networks. The EnKF does not require the explicit calculation of gradients or adjoints and we show this resolves the exploding and vanishing gradient problem. We analyze different parameter initializations, propose a dynamic change in ensembles and compare results to established methods.
000889208 536__ $$0G:(DE-HGF)POF3-511$$a511 - Computational Science and Mathematical Methods (POF3-511)$$cPOF3-511$$fPOF III$$x0
000889208 536__ $$0G:(DE-Juel1)HGF-SMHB-2013-2017$$aSMHB - Supercomputing and Modelling for the Human Brain (HGF-SMHB-2013-2017)$$cHGF-SMHB-2013-2017$$fSMHB$$x1
000889208 536__ $$0G:(DE-Juel1)CSD-SSD-20190612$$aCSD-SSD - Center for Simulation and Data Science (CSD) - School for Simulation and Data Science (SSD) (CSD-SSD-20190612)$$cCSD-SSD-20190612$$x2
000889208 536__ $$0G:(DE-Juel1)Helmholtz-SLNS$$aSLNS - SimLab Neuroscience (Helmholtz-SLNS)$$cHelmholtz-SLNS$$x3
000889208 536__ $$0G:(DE-Juel1)HDS-LEE-20190612$$aHDS LEE - Helmholtz School for Data Science in Life, Earth and Energy (HDS LEE) (HDS-LEE-20190612)$$cHDS-LEE-20190612$$x4
000889208 536__ $$0G:(DE-Juel1)PHD-NO-GRANT-20170405$$aPhD no Grant - Doktorand ohne besondere Förderung (PHD-NO-GRANT-20170405)$$cPHD-NO-GRANT-20170405$$x5
000889208 536__ $$0G:(DE-HGF)ZT-I-0003$$aHAF - Helmholtz Analytics Framework (ZT-I-0003)$$cZT-I-0003$$x6
000889208 7001_ $$0P:(DE-Juel1)129347$$aKrajsek, Kai$$b1$$ufzj
000889208 7001_ $$0P:(DE-Juel1)165859$$aDiaz, Sandra$$b2$$ufzj
000889208 7001_ $$0P:(DE-HGF)0$$aHerty, Michael$$b3
000889208 773__ $$a10.1007/978-3-030-64580-9_7
000889208 8564_ $$uhttps://juser.fz-juelich.de/record/889208/files/main.pdf$$yOpenAccess
000889208 909CO $$ooai:juser.fz-juelich.de:889208$$pdnbdelivery$$pdriver$$pVDB$$popen_access$$popenaire
000889208 9101_ $$0I:(DE-588b)5008462-8$$6P:(DE-Juel1)161462$$aForschungszentrum Jülich$$b0$$kFZJ
000889208 9101_ $$0I:(DE-588b)5008462-8$$6P:(DE-Juel1)129347$$aForschungszentrum Jülich$$b1$$kFZJ
000889208 9101_ $$0I:(DE-588b)5008462-8$$6P:(DE-Juel1)165859$$aForschungszentrum Jülich$$b2$$kFZJ
000889208 9131_ $$0G:(DE-HGF)POF3-511$$1G:(DE-HGF)POF3-510$$2G:(DE-HGF)POF3-500$$3G:(DE-HGF)POF3$$4G:(DE-HGF)POF$$aDE-HGF$$bKey Technologies$$lSupercomputing & Big Data$$vComputational Science and Mathematical Methods$$x0
000889208 9141_ $$y2020
000889208 915__ $$0StatID:(DE-HGF)0510$$2StatID$$aOpenAccess
000889208 920__ $$lyes
000889208 9201_ $$0I:(DE-Juel1)JSC-20090406$$kJSC$$lJülich Supercomputing Center$$x0
000889208 980__ $$acontrib
000889208 980__ $$aVDB
000889208 980__ $$aUNRESTRICTED
000889208 980__ $$acontb
000889208 980__ $$aI:(DE-Juel1)JSC-20090406
000889208 9801_ $$aFullTexts