Dissipative Deep Neural Dynamical Systems

In this paper, we provide sufficient conditions for dissipativity and local asymptotic stability of discrete-time dynamical systems parametrized by deep neural networks. We leverage the representation of neural networks as pointwise affine maps, thus exposing their local linear operators and making...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Drgona, Jan, Vasisht, Soumya, Tuor, Aaron, Vrabie, Draguna
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 Drgona, Jan
Vasisht, Soumya
Tuor, Aaron
Vrabie, Draguna
description In this paper, we provide sufficient conditions for dissipativity and local asymptotic stability of discrete-time dynamical systems parametrized by deep neural networks. We leverage the representation of neural networks as pointwise affine maps, thus exposing their local linear operators and making them accessible to classical system analytic and design methods. This allows us to "crack open the black box" of the neural dynamical system's behavior by evaluating their dissipativity, and estimating their stationary points and state-space partitioning. We relate the norms of these local linear operators to the energy stored in the dissipative system with supply rates represented by their aggregate bias terms. Empirically, we analyze the variance in dynamical behavior and eigenvalue spectra of these local linear operators with varying weight factorizations, activation functions, bias terms, and depths.
doi_str_mv 10.48550/arxiv.2011.13492
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2011_13492</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2011_13492</sourcerecordid><originalsourceid>FETCH-LOGICAL-a672-1d5006d885ae8de1abc5f39bd132ad5c23c257dcd095195c542ed12ebe96a2403</originalsourceid><addsrcrecordid>eNotzruOwjAQhWE3FCvgAbYiLUWCZ5wJcYkIe5EQFNBHE3uQLBEUxSzavD27QHX-6uhT6h10lpdEesH9b7hlqAEyMLnFNzWvQoyh42u4SVKJdMlOfno-J9Vw4Ta4vzoM8SptnKjRic9Rpq8dq-PH5rj-Srf7z-_1aptyscQUPGld-LIkltILcOPoZGzjwSB7cmgc0tI7ry2BJUc5igeURmzBmGszVrPn7cNad31ouR_qf3P9MJs7LC87Xw</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Dissipative Deep Neural Dynamical Systems</title><source>arXiv.org</source><creator>Drgona, Jan ; Vasisht, Soumya ; Tuor, Aaron ; Vrabie, Draguna</creator><creatorcontrib>Drgona, Jan ; Vasisht, Soumya ; Tuor, Aaron ; Vrabie, Draguna</creatorcontrib><description>In this paper, we provide sufficient conditions for dissipativity and local asymptotic stability of discrete-time dynamical systems parametrized by deep neural networks. We leverage the representation of neural networks as pointwise affine maps, thus exposing their local linear operators and making them accessible to classical system analytic and design methods. This allows us to "crack open the black box" of the neural dynamical system's behavior by evaluating their dissipativity, and estimating their stationary points and state-space partitioning. We relate the norms of these local linear operators to the energy stored in the dissipative system with supply rates represented by their aggregate bias terms. Empirically, we analyze the variance in dynamical behavior and eigenvalue spectra of these local linear operators with varying weight factorizations, activation functions, bias terms, and depths.</description><identifier>DOI: 10.48550/arxiv.2011.13492</identifier><language>eng</language><subject>Computer Science - Learning ; Computer Science - Neural and Evolutionary Computing</subject><creationdate>2020-11</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>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,780,885</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2011.13492$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2011.13492$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Drgona, Jan</creatorcontrib><creatorcontrib>Vasisht, Soumya</creatorcontrib><creatorcontrib>Tuor, Aaron</creatorcontrib><creatorcontrib>Vrabie, Draguna</creatorcontrib><title>Dissipative Deep Neural Dynamical Systems</title><description>In this paper, we provide sufficient conditions for dissipativity and local asymptotic stability of discrete-time dynamical systems parametrized by deep neural networks. We leverage the representation of neural networks as pointwise affine maps, thus exposing their local linear operators and making them accessible to classical system analytic and design methods. This allows us to "crack open the black box" of the neural dynamical system's behavior by evaluating their dissipativity, and estimating their stationary points and state-space partitioning. We relate the norms of these local linear operators to the energy stored in the dissipative system with supply rates represented by their aggregate bias terms. Empirically, we analyze the variance in dynamical behavior and eigenvalue spectra of these local linear operators with varying weight factorizations, activation functions, bias terms, and depths.</description><subject>Computer Science - Learning</subject><subject>Computer Science - Neural and Evolutionary Computing</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2020</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotzruOwjAQhWE3FCvgAbYiLUWCZ5wJcYkIe5EQFNBHE3uQLBEUxSzavD27QHX-6uhT6h10lpdEesH9b7hlqAEyMLnFNzWvQoyh42u4SVKJdMlOfno-J9Vw4Ta4vzoM8SptnKjRic9Rpq8dq-PH5rj-Srf7z-_1aptyscQUPGld-LIkltILcOPoZGzjwSB7cmgc0tI7ry2BJUc5igeURmzBmGszVrPn7cNad31ouR_qf3P9MJs7LC87Xw</recordid><startdate>20201126</startdate><enddate>20201126</enddate><creator>Drgona, Jan</creator><creator>Vasisht, Soumya</creator><creator>Tuor, Aaron</creator><creator>Vrabie, Draguna</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20201126</creationdate><title>Dissipative Deep Neural Dynamical Systems</title><author>Drgona, Jan ; Vasisht, Soumya ; Tuor, Aaron ; Vrabie, Draguna</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a672-1d5006d885ae8de1abc5f39bd132ad5c23c257dcd095195c542ed12ebe96a2403</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2020</creationdate><topic>Computer Science - Learning</topic><topic>Computer Science - Neural and Evolutionary Computing</topic><toplevel>online_resources</toplevel><creatorcontrib>Drgona, Jan</creatorcontrib><creatorcontrib>Vasisht, Soumya</creatorcontrib><creatorcontrib>Tuor, Aaron</creatorcontrib><creatorcontrib>Vrabie, Draguna</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Drgona, Jan</au><au>Vasisht, Soumya</au><au>Tuor, Aaron</au><au>Vrabie, Draguna</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Dissipative Deep Neural Dynamical Systems</atitle><date>2020-11-26</date><risdate>2020</risdate><abstract>In this paper, we provide sufficient conditions for dissipativity and local asymptotic stability of discrete-time dynamical systems parametrized by deep neural networks. We leverage the representation of neural networks as pointwise affine maps, thus exposing their local linear operators and making them accessible to classical system analytic and design methods. This allows us to "crack open the black box" of the neural dynamical system's behavior by evaluating their dissipativity, and estimating their stationary points and state-space partitioning. We relate the norms of these local linear operators to the energy stored in the dissipative system with supply rates represented by their aggregate bias terms. Empirically, we analyze the variance in dynamical behavior and eigenvalue spectra of these local linear operators with varying weight factorizations, activation functions, bias terms, and depths.</abstract><doi>10.48550/arxiv.2011.13492</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2011.13492
ispartof
issn
language eng
recordid cdi_arxiv_primary_2011_13492
source arXiv.org
subjects Computer Science - Learning
Computer Science - Neural and Evolutionary Computing
title Dissipative Deep Neural Dynamical Systems
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-27T00%3A18%3A50IST&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=Dissipative%20Deep%20Neural%20Dynamical%20Systems&rft.au=Drgona,%20Jan&rft.date=2020-11-26&rft_id=info:doi/10.48550/arxiv.2011.13492&rft_dat=%3Carxiv_GOX%3E2011_13492%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