STOCHASTIC GRADIENT LEARNING AND INSTABILITY: AN EXAMPLE

In this paper, we investigate real-time behavior of constant-gain stochastic gradient (SG) learning, using the Phelps model of monetary policy as a testing ground. We find that whereas the self-confirming equilibrium is stable under the mean dynamics in a very large region, real-time learning diverg...

Full description

Saved in:
Bibliographic Details
Published in:Macroeconomic dynamics 2016-04, Vol.20 (3), p.777-790
Main Authors: Slobodyan, Sergey, Bogomolova, Anna, Kolyuzhnov, Dmitri
Format: Article
Language:English
Subjects:
Algorithms
Approximation
Economic models
Economic statistics
Economic theory
Eigenvalues
Equilibrium
Macroeconomics
Monetary policy
Ordinary differential equations
Phillips curve
Studies
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
cited_by cdi_FETCH-LOGICAL-c350t-453038e59c28f96a3fcfcf796b593784f832e5f11878b88eac56f535527026ab3
cites cdi_FETCH-LOGICAL-c350t-453038e59c28f96a3fcfcf796b593784f832e5f11878b88eac56f535527026ab3
container_end_page 790
container_issue 3
container_start_page 777
container_title Macroeconomic dynamics
container_volume 20
creator Slobodyan, Sergey
Bogomolova, Anna
Kolyuzhnov, Dmitri
description In this paper, we investigate real-time behavior of constant-gain stochastic gradient (SG) learning, using the Phelps model of monetary policy as a testing ground. We find that whereas the self-confirming equilibrium is stable under the mean dynamics in a very large region, real-time learning diverges for all but the very smallest gain values. We employ a stochastic Lyapunov function approach to demonstrate that the SG mean dynamics is easily destabilized by the noise associated with real-time learning, because its Jacobian contains stable but very small eigenvalues. We also express caution on usage of perpetual learning algorithms with such small eigenvalues, as the real-time dynamics might diverge from the equilibrium that is stable under the mean dynamics.
doi_str_mv 10.1017/S1365100514000583
format article
fullrecord <record><control><sourceid>proquest_cross</sourceid><recordid>TN_cdi_proquest_journals_1777431649</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><cupid>10_1017_S1365100514000583</cupid><sourcerecordid>4005627191</sourcerecordid><originalsourceid>FETCH-LOGICAL-c350t-453038e59c28f96a3fcfcf796b593784f832e5f11878b88eac56f535527026ab3</originalsourceid><addsrcrecordid>eNp1UEtLw0AQXkTBWv0B3gKeozvZ7MvbmqbpQkyliaCnkMRdabGmbtqD_94t7UEQGZgZ5nsMfAhdA74FDPyuBMIoYEwhxr4LcoJGEDMZCkzZqd89HO7xc3QxDCuMgZFIjpAoq3kyU2WlkyBbqIlOiyrIU7UodJEFqpgEuigr9aBzXb3e-0OQvqjHpzy9RGe2-RjM1XGO0fM0rZJZmM8znag87AjF2zCmBBNhqOwiYSVriO18cclaKgkXsRUkMtQCCC5aIUzTUWYpoTTiOGJNS8bo5uC7cf3XzgzbetXv3Kd_WQPnPCbAYulZcGB1rh8GZ2y9cct1475rwPU-oPpPQF5Djppm3brl27v5Zf2v6gcUTV-3</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>1777431649</pqid></control><display><type>article</type><title>STOCHASTIC GRADIENT LEARNING AND INSTABILITY: AN EXAMPLE</title><source>EconLit s plnými texty</source><source>ABI/INFORM Global (ProQuest)</source><source>Cambridge University Press</source><creator>Slobodyan, Sergey ; Bogomolova, Anna ; Kolyuzhnov, Dmitri</creator><creatorcontrib>Slobodyan, Sergey ; Bogomolova, Anna ; Kolyuzhnov, Dmitri</creatorcontrib><description>In this paper, we investigate real-time behavior of constant-gain stochastic gradient (SG) learning, using the Phelps model of monetary policy as a testing ground. We find that whereas the self-confirming equilibrium is stable under the mean dynamics in a very large region, real-time learning diverges for all but the very smallest gain values. We employ a stochastic Lyapunov function approach to demonstrate that the SG mean dynamics is easily destabilized by the noise associated with real-time learning, because its Jacobian contains stable but very small eigenvalues. We also express caution on usage of perpetual learning algorithms with such small eigenvalues, as the real-time dynamics might diverge from the equilibrium that is stable under the mean dynamics.</description><identifier>ISSN: 1365-1005</identifier><identifier>EISSN: 1469-8056</identifier><identifier>DOI: 10.1017/S1365100514000583</identifier><language>eng</language><publisher>New York, USA: Cambridge University Press</publisher><subject>Algorithms ; Approximation ; Economic models ; Economic statistics ; Economic theory ; Eigenvalues ; Equilibrium ; Macroeconomics ; Monetary policy ; Ordinary differential equations ; Phillips curve ; Studies</subject><ispartof>Macroeconomic dynamics, 2016-04, Vol.20 (3), p.777-790</ispartof><rights>Copyright © Cambridge University Press 2016</rights><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c350t-453038e59c28f96a3fcfcf796b593784f832e5f11878b88eac56f535527026ab3</citedby><cites>FETCH-LOGICAL-c350t-453038e59c28f96a3fcfcf796b593784f832e5f11878b88eac56f535527026ab3</cites></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktopdf>$$Uhttps://www.proquest.com/docview/1777431649/fulltextPDF?pq-origsite=primo$$EPDF$$P50$$Gproquest$$H</linktopdf><linktohtml>$$Uhttps://www.proquest.com/docview/1777431649?pq-origsite=primo$$EHTML$$P50$$Gproquest$$H</linktohtml><link.rule.ids>314,780,784,11688,27924,27925,36060,44363,72960,74895</link.rule.ids></links><search><creatorcontrib>Slobodyan, Sergey</creatorcontrib><creatorcontrib>Bogomolova, Anna</creatorcontrib><creatorcontrib>Kolyuzhnov, Dmitri</creatorcontrib><title>STOCHASTIC GRADIENT LEARNING AND INSTABILITY: AN EXAMPLE</title><title>Macroeconomic dynamics</title><addtitle>Macroecon. Dynam</addtitle><description>In this paper, we investigate real-time behavior of constant-gain stochastic gradient (SG) learning, using the Phelps model of monetary policy as a testing ground. We find that whereas the self-confirming equilibrium is stable under the mean dynamics in a very large region, real-time learning diverges for all but the very smallest gain values. We employ a stochastic Lyapunov function approach to demonstrate that the SG mean dynamics is easily destabilized by the noise associated with real-time learning, because its Jacobian contains stable but very small eigenvalues. We also express caution on usage of perpetual learning algorithms with such small eigenvalues, as the real-time dynamics might diverge from the equilibrium that is stable under the mean dynamics.</description><subject>Algorithms</subject><subject>Approximation</subject><subject>Economic models</subject><subject>Economic statistics</subject><subject>Economic theory</subject><subject>Eigenvalues</subject><subject>Equilibrium</subject><subject>Macroeconomics</subject><subject>Monetary policy</subject><subject>Ordinary differential equations</subject><subject>Phillips curve</subject><subject>Studies</subject><issn>1365-1005</issn><issn>1469-8056</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2016</creationdate><recordtype>article</recordtype><sourceid>M0C</sourceid><recordid>eNp1UEtLw0AQXkTBWv0B3gKeozvZ7MvbmqbpQkyliaCnkMRdabGmbtqD_94t7UEQGZgZ5nsMfAhdA74FDPyuBMIoYEwhxr4LcoJGEDMZCkzZqd89HO7xc3QxDCuMgZFIjpAoq3kyU2WlkyBbqIlOiyrIU7UodJEFqpgEuigr9aBzXb3e-0OQvqjHpzy9RGe2-RjM1XGO0fM0rZJZmM8znag87AjF2zCmBBNhqOwiYSVriO18cclaKgkXsRUkMtQCCC5aIUzTUWYpoTTiOGJNS8bo5uC7cf3XzgzbetXv3Kd_WQPnPCbAYulZcGB1rh8GZ2y9cct1475rwPU-oPpPQF5Djppm3brl27v5Zf2v6gcUTV-3</recordid><startdate>201604</startdate><enddate>201604</enddate><creator>Slobodyan, Sergey</creator><creator>Bogomolova, Anna</creator><creator>Kolyuzhnov, Dmitri</creator><general>Cambridge University Press</general><scope>AAYXX</scope><scope>CITATION</scope><scope>0U~</scope><scope>1-H</scope><scope>3V.</scope><scope>7WY</scope><scope>7WZ</scope><scope>7XB</scope><scope>87Z</scope><scope>8FK</scope><scope>8FL</scope><scope>8G5</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BEZIV</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>FRNLG</scope><scope>F~G</scope><scope>GNUQQ</scope><scope>GUQSH</scope><scope>K60</scope><scope>K6~</scope><scope>L.-</scope><scope>L.0</scope><scope>M0C</scope><scope>M2O</scope><scope>MBDVC</scope><scope>PQBIZ</scope><scope>PQBZA</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>Q9U</scope></search><sort><creationdate>201604</creationdate><title>STOCHASTIC GRADIENT LEARNING AND INSTABILITY: AN EXAMPLE</title><author>Slobodyan, Sergey ; Bogomolova, Anna ; Kolyuzhnov, Dmitri</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c350t-453038e59c28f96a3fcfcf796b593784f832e5f11878b88eac56f535527026ab3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2016</creationdate><topic>Algorithms</topic><topic>Approximation</topic><topic>Economic models</topic><topic>Economic statistics</topic><topic>Economic theory</topic><topic>Eigenvalues</topic><topic>Equilibrium</topic><topic>Macroeconomics</topic><topic>Monetary policy</topic><topic>Ordinary differential equations</topic><topic>Phillips curve</topic><topic>Studies</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Slobodyan, Sergey</creatorcontrib><creatorcontrib>Bogomolova, Anna</creatorcontrib><creatorcontrib>Kolyuzhnov, Dmitri</creatorcontrib><collection>CrossRef</collection><collection>Global News &amp; ABI/Inform Professional</collection><collection>Trade PRO</collection><collection>ProQuest Central (Corporate)</collection><collection>ABI商业信息数据库</collection><collection>ABI/INFORM Global (PDF only)</collection><collection>ProQuest Central (purchase pre-March 2016)</collection><collection>ABI/INFORM Collection</collection><collection>ProQuest Central (Alumni) (purchase pre-March 2016)</collection><collection>ABI/INFORM Collection (Alumni Edition)</collection><collection>Research Library (Alumni Edition)</collection><collection>ProQuest Central (Alumni)</collection><collection>ProQuest Central</collection><collection>ProQuest Central Essentials</collection><collection>AUTh Library subscriptions: ProQuest Central</collection><collection>ProQuest Business Premium Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central</collection><collection>Business Premium Collection (Alumni)</collection><collection>ABI/INFORM Global (Corporate)</collection><collection>ProQuest Central Student</collection><collection>Research Library Prep</collection><collection>ProQuest Business Collection (Alumni Edition)</collection><collection>ProQuest Business Collection</collection><collection>ABI/INFORM Professional Advanced</collection><collection>ABI/INFORM Professional Standard</collection><collection>ABI/INFORM Global (ProQuest)</collection><collection>ProQuest research library</collection><collection>Research Library (Corporate)</collection><collection>One Business (ProQuest)</collection><collection>ProQuest One Business (Alumni)</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central Basic</collection><jtitle>Macroeconomic dynamics</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Slobodyan, Sergey</au><au>Bogomolova, Anna</au><au>Kolyuzhnov, Dmitri</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>STOCHASTIC GRADIENT LEARNING AND INSTABILITY: AN EXAMPLE</atitle><jtitle>Macroeconomic dynamics</jtitle><addtitle>Macroecon. Dynam</addtitle><date>2016-04</date><risdate>2016</risdate><volume>20</volume><issue>3</issue><spage>777</spage><epage>790</epage><pages>777-790</pages><issn>1365-1005</issn><eissn>1469-8056</eissn><abstract>In this paper, we investigate real-time behavior of constant-gain stochastic gradient (SG) learning, using the Phelps model of monetary policy as a testing ground. We find that whereas the self-confirming equilibrium is stable under the mean dynamics in a very large region, real-time learning diverges for all but the very smallest gain values. We employ a stochastic Lyapunov function approach to demonstrate that the SG mean dynamics is easily destabilized by the noise associated with real-time learning, because its Jacobian contains stable but very small eigenvalues. We also express caution on usage of perpetual learning algorithms with such small eigenvalues, as the real-time dynamics might diverge from the equilibrium that is stable under the mean dynamics.</abstract><cop>New York, USA</cop><pub>Cambridge University Press</pub><doi>10.1017/S1365100514000583</doi><tpages>14</tpages></addata></record>
fulltext fulltext
identifier ISSN: 1365-1005
ispartof Macroeconomic dynamics, 2016-04, Vol.20 (3), p.777-790
issn 1365-1005
1469-8056
language eng
recordid cdi_proquest_journals_1777431649
source EconLit s plnými texty; ABI/INFORM Global (ProQuest); Cambridge University Press
subjects Algorithms
Approximation
Economic models
Economic statistics
Economic theory
Eigenvalues
Equilibrium
Macroeconomics
Monetary policy
Ordinary differential equations
Phillips curve
Studies
title STOCHASTIC GRADIENT LEARNING AND INSTABILITY: AN EXAMPLE
url http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-25T09%3A02%3A57IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_cross&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=STOCHASTIC%20GRADIENT%20LEARNING%20AND%20INSTABILITY:%20AN%20EXAMPLE&rft.jtitle=Macroeconomic%20dynamics&rft.au=Slobodyan,%20Sergey&rft.date=2016-04&rft.volume=20&rft.issue=3&rft.spage=777&rft.epage=790&rft.pages=777-790&rft.issn=1365-1005&rft.eissn=1469-8056&rft_id=info:doi/10.1017/S1365100514000583&rft_dat=%3Cproquest_cross%3E4005627191%3C/proquest_cross%3E%3Cgrp_id%3Ecdi_FETCH-LOGICAL-c350t-453038e59c28f96a3fcfcf796b593784f832e5f11878b88eac56f535527026ab3%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_pqid=1777431649&rft_id=info:pmid/&rft_cupid=10_1017_S1365100514000583&rfr_iscdi=true