Please use this identifier to cite or link to this item:
https://hdl.handle.net/10419/19040
Full metadata record
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Evans, George W. | en |
dc.contributor.author | Honkapohja, Seppo | en |
dc.contributor.author | Williams, Noah | en |
dc.date.accessioned | 2009-01-28T15:54:38Z | - |
dc.date.available | 2009-01-28T15:54:38Z | - |
dc.date.issued | 2005 | - |
dc.identifier.uri | http://hdl.handle.net/10419/19040 | - |
dc.description.abstract | We study the properties of generalized stochastic gradient (GSG) learning in forward-lookingmodels. We examine how the conditions for stability of standard stochastic gradient (SG)learning both differ from and are related to E-stability, which governs stability under leastsquares learning. SG algorithms are sensitive to units of measurement and we show that thereis a transformation of variables for which E-stability governs SG stability. GSG algorithmswith constant gain have a deeper justification in terms of parameter drift, robustness and risksensitivity. | en |
dc.language.iso | eng | en |
dc.publisher | |aCenter for Economic Studies and ifo Institute (CESifo) |cMunich | en |
dc.relation.ispartofseries | |aCESifo Working Paper |x1576 | en |
dc.subject.jel | C65 | en |
dc.subject.jel | C62 | en |
dc.subject.jel | E17 | en |
dc.subject.jel | E10 | en |
dc.subject.jel | D83 | en |
dc.subject.ddc | 330 | en |
dc.subject.keyword | adaptive learning | en |
dc.subject.keyword | E-stability | en |
dc.subject.keyword | recursive least squares | en |
dc.subject.keyword | robust estimation | en |
dc.subject.stw | Rationale Erwartung | en |
dc.subject.stw | Lernprozess | en |
dc.subject.stw | Prognoseverfahren | en |
dc.subject.stw | Gleichgewichtsstabilität | en |
dc.subject.stw | Theorie | en |
dc.title | Generalized stochastic gradient learning | - |
dc.type | Working Paper | en |
dc.identifier.ppn | 503712469 | en |
dc.rights | http://www.econstor.eu/dspace/Nutzungsbedingungen | en |
Files in This Item:
Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.