Weak Convergence Analysis of Online Neural Actor-Critic Algorithms

We prove that a single-layer neural network trained with the online actor critic algorithm converges in distribution to a random ordinary differential equation (ODE) as the number of hidden units and the number of training steps $\rightarrow \infty$. In the online actor-critic algorithm, the distrib...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Lam, Samuel Chun-Hei, Sirignano, Justin, Wang, Ziheng
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Lam, Samuel Chun-Hei
Sirignano, Justin
Wang, Ziheng
description We prove that a single-layer neural network trained with the online actor critic algorithm converges in distribution to a random ordinary differential equation (ODE) as the number of hidden units and the number of training steps $\rightarrow \infty$. In the online actor-critic algorithm, the distribution of the data samples dynamically changes as the model is updated, which is a key challenge for any convergence analysis. We establish the geometric ergodicity of the data samples under a fixed actor policy. Then, using a Poisson equation, we prove that the fluctuations of the model updates around the limit distribution due to the randomly-arriving data samples vanish as the number of parameter updates $\rightarrow \infty$. Using the Poisson equation and weak convergence techniques, we prove that the actor neural network and critic neural network converge to the solutions of a system of ODEs with random initial conditions. Analysis of the limit ODE shows that the limit critic network will converge to the true value function, which will provide the actor an asymptotically unbiased estimate of the policy gradient. We then prove that the limit actor network will converge to a stationary point.
doi_str_mv 10.48550/arxiv.2403.16825
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2403_16825</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2403_16825</sourcerecordid><originalsourceid>FETCH-LOGICAL-a675-d653705326a9cf4c3e2511c44b89107bddff89b2a13b7538d0096889665a7bd43</originalsourceid><addsrcrecordid>eNotz8tOwzAUBFBvWKDCB7DCP5Dg13XsZYh4SRXdVGIZ3Th2a9V1kFMq-veUwmpGGmmkQ8gdZ7UyAOwBy3c81kIxWXNtBFyTxw-PO9pN-ejLxmfnaZsxneY40ynQVU4xe_ruvwom2rrDVKquxEN0tE2b6dy2-_mGXAVMs7_9zwVZPz-tu9dquXp569plhbqBatQgGwZSaLQuKCe9AM6dUoOxnDXDOIZg7CCQy6EBaUbGrDbGag14XpVckPu_2wui_yxxj-XU_2L6C0b-AEXIQ3I</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Weak Convergence Analysis of Online Neural Actor-Critic Algorithms</title><source>arXiv.org</source><creator>Lam, Samuel Chun-Hei ; Sirignano, Justin ; Wang, Ziheng</creator><creatorcontrib>Lam, Samuel Chun-Hei ; Sirignano, Justin ; Wang, Ziheng</creatorcontrib><description>We prove that a single-layer neural network trained with the online actor critic algorithm converges in distribution to a random ordinary differential equation (ODE) as the number of hidden units and the number of training steps $\rightarrow \infty$. In the online actor-critic algorithm, the distribution of the data samples dynamically changes as the model is updated, which is a key challenge for any convergence analysis. We establish the geometric ergodicity of the data samples under a fixed actor policy. Then, using a Poisson equation, we prove that the fluctuations of the model updates around the limit distribution due to the randomly-arriving data samples vanish as the number of parameter updates $\rightarrow \infty$. Using the Poisson equation and weak convergence techniques, we prove that the actor neural network and critic neural network converge to the solutions of a system of ODEs with random initial conditions. Analysis of the limit ODE shows that the limit critic network will converge to the true value function, which will provide the actor an asymptotically unbiased estimate of the policy gradient. We then prove that the limit actor network will converge to a stationary point.</description><identifier>DOI: 10.48550/arxiv.2403.16825</identifier><language>eng</language><subject>Computer Science - Learning ; Mathematics - Optimization and Control ; Mathematics - Probability ; Statistics - Machine Learning</subject><creationdate>2024-03</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,777,882</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2403.16825$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2403.16825$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Lam, Samuel Chun-Hei</creatorcontrib><creatorcontrib>Sirignano, Justin</creatorcontrib><creatorcontrib>Wang, Ziheng</creatorcontrib><title>Weak Convergence Analysis of Online Neural Actor-Critic Algorithms</title><description>We prove that a single-layer neural network trained with the online actor critic algorithm converges in distribution to a random ordinary differential equation (ODE) as the number of hidden units and the number of training steps $\rightarrow \infty$. In the online actor-critic algorithm, the distribution of the data samples dynamically changes as the model is updated, which is a key challenge for any convergence analysis. We establish the geometric ergodicity of the data samples under a fixed actor policy. Then, using a Poisson equation, we prove that the fluctuations of the model updates around the limit distribution due to the randomly-arriving data samples vanish as the number of parameter updates $\rightarrow \infty$. Using the Poisson equation and weak convergence techniques, we prove that the actor neural network and critic neural network converge to the solutions of a system of ODEs with random initial conditions. Analysis of the limit ODE shows that the limit critic network will converge to the true value function, which will provide the actor an asymptotically unbiased estimate of the policy gradient. We then prove that the limit actor network will converge to a stationary point.</description><subject>Computer Science - Learning</subject><subject>Mathematics - Optimization and Control</subject><subject>Mathematics - Probability</subject><subject>Statistics - Machine Learning</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotz8tOwzAUBFBvWKDCB7DCP5Dg13XsZYh4SRXdVGIZ3Th2a9V1kFMq-veUwmpGGmmkQ8gdZ7UyAOwBy3c81kIxWXNtBFyTxw-PO9pN-ejLxmfnaZsxneY40ynQVU4xe_ruvwom2rrDVKquxEN0tE2b6dy2-_mGXAVMs7_9zwVZPz-tu9dquXp569plhbqBatQgGwZSaLQuKCe9AM6dUoOxnDXDOIZg7CCQy6EBaUbGrDbGag14XpVckPu_2wui_yxxj-XU_2L6C0b-AEXIQ3I</recordid><startdate>20240325</startdate><enddate>20240325</enddate><creator>Lam, Samuel Chun-Hei</creator><creator>Sirignano, Justin</creator><creator>Wang, Ziheng</creator><scope>AKY</scope><scope>AKZ</scope><scope>EPD</scope><scope>GOX</scope></search><sort><creationdate>20240325</creationdate><title>Weak Convergence Analysis of Online Neural Actor-Critic Algorithms</title><author>Lam, Samuel Chun-Hei ; Sirignano, Justin ; Wang, Ziheng</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a675-d653705326a9cf4c3e2511c44b89107bddff89b2a13b7538d0096889665a7bd43</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Computer Science - Learning</topic><topic>Mathematics - Optimization and Control</topic><topic>Mathematics - Probability</topic><topic>Statistics - Machine Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Lam, Samuel Chun-Hei</creatorcontrib><creatorcontrib>Sirignano, Justin</creatorcontrib><creatorcontrib>Wang, Ziheng</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv Mathematics</collection><collection>arXiv Statistics</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Lam, Samuel Chun-Hei</au><au>Sirignano, Justin</au><au>Wang, Ziheng</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Weak Convergence Analysis of Online Neural Actor-Critic Algorithms</atitle><date>2024-03-25</date><risdate>2024</risdate><abstract>We prove that a single-layer neural network trained with the online actor critic algorithm converges in distribution to a random ordinary differential equation (ODE) as the number of hidden units and the number of training steps $\rightarrow \infty$. In the online actor-critic algorithm, the distribution of the data samples dynamically changes as the model is updated, which is a key challenge for any convergence analysis. We establish the geometric ergodicity of the data samples under a fixed actor policy. Then, using a Poisson equation, we prove that the fluctuations of the model updates around the limit distribution due to the randomly-arriving data samples vanish as the number of parameter updates $\rightarrow \infty$. Using the Poisson equation and weak convergence techniques, we prove that the actor neural network and critic neural network converge to the solutions of a system of ODEs with random initial conditions. Analysis of the limit ODE shows that the limit critic network will converge to the true value function, which will provide the actor an asymptotically unbiased estimate of the policy gradient. We then prove that the limit actor network will converge to a stationary point.</abstract><doi>10.48550/arxiv.2403.16825</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2403.16825
ispartof
issn
language eng
recordid cdi_arxiv_primary_2403_16825
source arXiv.org
subjects Computer Science - Learning
Mathematics - Optimization and Control
Mathematics - Probability
Statistics - Machine Learning
title Weak Convergence Analysis of Online Neural Actor-Critic Algorithms
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-20T00%3A16%3A59IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Weak%20Convergence%20Analysis%20of%20Online%20Neural%20Actor-Critic%20Algorithms&rft.au=Lam,%20Samuel%20Chun-Hei&rft.date=2024-03-25&rft_id=info:doi/10.48550/arxiv.2403.16825&rft_dat=%3Carxiv_GOX%3E2403_16825%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true