Accuracy analysis of uncertain variational problems with analytical and machine learning methods

Tässä tutkielmassa verrataan analyyttisien menetelmien ja koneoppimismallien toimivuutta epätarkkuudesta johtuvien virheiden kontrolloinnissa. Tarkasteltavana matemaattisena esimerkkiongelmana käytetään lineaarista variaatio-ongelmaa. Tuloksena havaitaan, että neuroverkot toimivat hyvin ja ovat käyt...

Täydet tiedot

Bibliografiset tiedot
Päätekijä: Halonen, Vilho
Muut tekijät: Matemaattis-luonnontieteellinen tiedekunta, Faculty of Sciences, Matematiikan ja tilastotieteen laitos, Department of Mathematics and Statistics, Jyväskylän yliopisto, University of Jyväskylä
Aineistotyyppi: Pro gradu
Kieli:eng
Julkaistu: 2021
Aiheet:
Linkit: https://jyx.jyu.fi/handle/123456789/78718
_version_ 1828193066852286464
author Halonen, Vilho
author2 Matemaattis-luonnontieteellinen tiedekunta Faculty of Sciences Matematiikan ja tilastotieteen laitos Department of Mathematics and Statistics Jyväskylän yliopisto University of Jyväskylä
author_facet Halonen, Vilho Matemaattis-luonnontieteellinen tiedekunta Faculty of Sciences Matematiikan ja tilastotieteen laitos Department of Mathematics and Statistics Jyväskylän yliopisto University of Jyväskylä Halonen, Vilho Matemaattis-luonnontieteellinen tiedekunta Faculty of Sciences Matematiikan ja tilastotieteen laitos Department of Mathematics and Statistics Jyväskylän yliopisto University of Jyväskylä
author_sort Halonen, Vilho
datasource_str_mv jyx
description Tässä tutkielmassa verrataan analyyttisien menetelmien ja koneoppimismallien toimivuutta epätarkkuudesta johtuvien virheiden kontrolloinnissa. Tarkasteltavana matemaattisena esimerkkiongelmana käytetään lineaarista variaatio-ongelmaa. Tuloksena havaitaan, että neuroverkot toimivat hyvin ja ovat käytäntöön mahdollisesti soveltuva keino tehdä virhearviointia. Monille osittaisdifferentiaaliyhtälöille on johdettu analyyttisia kontrollointikeinoja viime vuosikymmenien aikana (katso [1], [2]). Ensimmäiset luvut käytämme analyyttisien virhearvioiden todistamiseen tunnettujen analyysin työkalujen avulla tarkasteltavalle variaatio-ongelmalle. Virhearvioita testataan numeerisesti ja huomataan, että vaikka analyyttiset rajat ovat varmoja ja halpoja laskennallisesti, ne ovat monesti toivottua epätarkempia. Tutkielman toisessa osiossa luodaan koneoppimismalleja, joilla pyritään arvioimaan tarkalleen epätarkkuuden aiheuttamaa virhettä. Valittu koneoppimismalli on neuroverkko. Mallien kouluttamiseen käytetty data luodaan itse numeerisilla menetelmillä. Viimeisessä luvussa verrataan analyyttisien metodien ja luotujen neuroverkkojen toimivuutta. Vertailussa käytetään koulutusdatasta eroavaa generoitua dataa jolle lasketaan analyyttiset rajat, numeeriset approksimaatiot ja neuroverkkojen tulokset. Havaitaan, että neuroverkot suoriutuvat tehtävästä niin hyvin, että voidaan sanoa niiden olevan kilpailullisia analyyttisien metodien kanssa. Jos vastaavia koneoppimismalleja pystytään luomaan vaikeammille moniulotteisille ongelmille, tämä menetelmä voi osoittautua varsin hyödylliseksi simuloinnissa ja insinöörityössä. In this thesis we compare the performance of analytical methods and neural networks trained with numerically produced data in controlling uncertainty errors of a linear variational problem. We find that neural networks perform well and are feasible to use in practical computations in place of analytical control methods. Analytical methods for controlling uncertainty errors have been derived for various differential problems (see [1], [2]) in recent decades. The first chapters are devoted to deriving by known methods analytical error bounds for the linear variational problem which we will study. These error bounds are numerically tested and we find that the bounds while they are guaranteed and cheap to compute are not always as sharp as an engineer might hope. The second part of this thesis consists of creating machine learning models with the goal of approximating the exact error caused by uncertainty in our mathematical model. The chosen type of machine learning model is a deep neural network. The training data used for training the models is generated by numerical computations. In the final chapter we compare the performance of the analytical methods and machine learning models and we conclude that neural networks can be competitive in this task. If such models are made and found to work for more complicated nonlinear PDE’s this method could prove very useful in computer simulations and engineering.
first_indexed 2021-11-19T21:00:45Z
format Pro gradu
free_online_boolean 1
fullrecord [{"key": "dc.contributor.advisor", "value": "Repin, Sergey", "language": "", "element": "contributor", "qualifier": "advisor", "schema": "dc"}, {"key": "dc.contributor.advisor", "value": "P\u00f6l\u00f6nen, Ilkka", "language": "", "element": "contributor", "qualifier": "advisor", "schema": "dc"}, {"key": "dc.contributor.advisor", "value": "Julin, Vesa", "language": "", "element": "contributor", "qualifier": "advisor", "schema": "dc"}, {"key": "dc.contributor.author", "value": "Halonen, Vilho", "language": "", "element": "contributor", "qualifier": "author", "schema": "dc"}, {"key": "dc.date.accessioned", "value": "2021-11-19T08:00:49Z", "language": null, "element": "date", "qualifier": "accessioned", "schema": "dc"}, {"key": "dc.date.available", "value": "2021-11-19T08:00:49Z", "language": null, "element": "date", "qualifier": "available", "schema": "dc"}, {"key": "dc.date.issued", "value": "2021", "language": "", "element": "date", "qualifier": "issued", "schema": "dc"}, {"key": "dc.identifier.uri", "value": "https://jyx.jyu.fi/handle/123456789/78718", "language": null, "element": "identifier", "qualifier": "uri", "schema": "dc"}, {"key": "dc.description.abstract", "value": "T\u00e4ss\u00e4 tutkielmassa verrataan analyyttisien menetelmien ja koneoppimismallien toimivuutta ep\u00e4tarkkuudesta johtuvien virheiden kontrolloinnissa. Tarkasteltavana matemaattisena esimerkkiongelmana k\u00e4ytet\u00e4\u00e4n lineaarista variaatio-ongelmaa. Tuloksena havaitaan, ett\u00e4 neuroverkot toimivat hyvin ja ovat k\u00e4yt\u00e4nt\u00f6\u00f6n mahdollisesti soveltuva keino tehd\u00e4 virhearviointia.\nMonille osittaisdifferentiaaliyht\u00e4l\u00f6ille on johdettu analyyttisia kontrollointikeinoja viime vuosikymmenien aikana (katso [1], [2]). Ensimm\u00e4iset luvut k\u00e4yt\u00e4mme analyyttisien virhearvioiden todistamiseen tunnettujen analyysin ty\u00f6kalujen avulla tarkasteltavalle variaatio-ongelmalle. Virhearvioita testataan numeerisesti ja huomataan, ett\u00e4 vaikka analyyttiset rajat ovat varmoja ja halpoja laskennallisesti, ne ovat monesti toivottua ep\u00e4tarkempia.\nTutkielman toisessa osiossa luodaan koneoppimismalleja, joilla pyrit\u00e4\u00e4n arvioimaan tarkalleen ep\u00e4tarkkuuden aiheuttamaa virhett\u00e4. Valittu koneoppimismalli on neuroverkko. Mallien kouluttamiseen k\u00e4ytetty data luodaan itse numeerisilla menetelmill\u00e4.\nViimeisess\u00e4 luvussa verrataan analyyttisien metodien ja luotujen neuroverkkojen toimivuutta. Vertailussa k\u00e4ytet\u00e4\u00e4n koulutusdatasta eroavaa generoitua dataa jolle lasketaan analyyttiset rajat, numeeriset approksimaatiot ja neuroverkkojen tulokset. Havaitaan, ett\u00e4 neuroverkot suoriutuvat teht\u00e4v\u00e4st\u00e4 niin hyvin, ett\u00e4 voidaan sanoa niiden olevan kilpailullisia analyyttisien metodien kanssa. Jos vastaavia koneoppimismalleja pystyt\u00e4\u00e4n luomaan vaikeammille moniulotteisille ongelmille, t\u00e4m\u00e4 menetelm\u00e4 voi osoittautua varsin hy\u00f6dylliseksi simuloinnissa ja insin\u00f6\u00f6rity\u00f6ss\u00e4.", "language": "fi", "element": "description", "qualifier": "abstract", "schema": "dc"}, {"key": "dc.description.abstract", "value": "In this thesis we compare the performance of analytical methods and neural networks trained with numerically produced data in controlling uncertainty errors of a linear variational problem. We find that neural networks perform well and are feasible to use in practical computations in place of analytical control methods.\nAnalytical methods for controlling uncertainty errors have been derived for various differential problems (see [1], [2]) in recent decades. The first chapters are devoted to deriving by known methods analytical error bounds for the linear variational problem which we will study. These error bounds are numerically tested and we find that the bounds while they are guaranteed and cheap to compute are not always as sharp as an engineer might hope.\nThe second part of this thesis consists of creating machine learning models with the goal of approximating the exact error caused by uncertainty in our mathematical model. The chosen type of machine learning model is a deep neural network. The training data used for training the models is generated by numerical computations.\nIn the final chapter we compare the performance of the analytical methods and machine learning models and we conclude that neural networks can be competitive in this task. If such models are made and found to work for more complicated nonlinear PDE\u2019s this method could prove very useful in computer simulations and engineering.", "language": "en", "element": "description", "qualifier": "abstract", "schema": "dc"}, {"key": "dc.description.provenance", "value": "Submitted by Paivi Vuorio (paelvuor@jyu.fi) on 2021-11-19T08:00:49Z\nNo. of bitstreams: 0", "language": "en", "element": "description", "qualifier": "provenance", "schema": "dc"}, {"key": "dc.description.provenance", "value": "Made available in DSpace on 2021-11-19T08:00:49Z (GMT). No. of bitstreams: 0\n Previous issue date: 2021", "language": "en", "element": "description", "qualifier": "provenance", "schema": "dc"}, {"key": "dc.format.extent", "value": "63", "language": "", "element": "format", "qualifier": "extent", "schema": "dc"}, {"key": "dc.format.mimetype", "value": "application/pdf", "language": null, "element": "format", "qualifier": "mimetype", "schema": "dc"}, {"key": "dc.language.iso", "value": "eng", "language": null, "element": "language", "qualifier": "iso", "schema": "dc"}, {"key": "dc.rights", "value": "In Copyright", "language": "en", "element": "rights", "qualifier": null, "schema": "dc"}, {"key": "dc.title", "value": "Accuracy analysis of uncertain variational problems with analytical and machine learning methods", "language": "", "element": "title", "qualifier": null, "schema": "dc"}, {"key": "dc.type", "value": "master thesis", "language": null, "element": "type", "qualifier": null, "schema": "dc"}, {"key": "dc.identifier.urn", "value": "URN:NBN:fi:jyu-202111195729", "language": "", "element": "identifier", "qualifier": "urn", "schema": "dc"}, {"key": "dc.type.ontasot", "value": "Pro gradu -tutkielma", "language": "fi", "element": "type", "qualifier": "ontasot", "schema": "dc"}, {"key": "dc.type.ontasot", "value": "Master\u2019s thesis", "language": "en", "element": "type", "qualifier": "ontasot", "schema": "dc"}, {"key": "dc.contributor.faculty", "value": "Matemaattis-luonnontieteellinen tiedekunta", "language": "fi", "element": "contributor", "qualifier": "faculty", "schema": "dc"}, {"key": "dc.contributor.faculty", "value": "Faculty of Sciences", "language": "en", "element": "contributor", "qualifier": "faculty", "schema": "dc"}, {"key": "dc.contributor.department", "value": "Matematiikan ja tilastotieteen laitos", "language": "fi", "element": "contributor", "qualifier": "department", "schema": "dc"}, {"key": "dc.contributor.department", "value": "Department of Mathematics and Statistics", "language": "en", "element": "contributor", "qualifier": "department", "schema": "dc"}, {"key": "dc.contributor.organization", "value": "Jyv\u00e4skyl\u00e4n yliopisto", "language": "fi", "element": "contributor", "qualifier": "organization", "schema": "dc"}, {"key": "dc.contributor.organization", "value": "University of Jyv\u00e4skyl\u00e4", "language": "en", "element": "contributor", "qualifier": "organization", "schema": "dc"}, {"key": "dc.subject.discipline", "value": "Matematiikka", "language": "fi", "element": "subject", "qualifier": "discipline", "schema": "dc"}, {"key": "dc.subject.discipline", "value": "Mathematics", "language": "en", "element": "subject", "qualifier": "discipline", "schema": "dc"}, {"key": "yvv.contractresearch.funding", "value": "0", "language": "", "element": "contractresearch", "qualifier": "funding", "schema": "yvv"}, {"key": "dc.type.coar", "value": "http://purl.org/coar/resource_type/c_bdcc", "language": null, "element": "type", "qualifier": "coar", "schema": "dc"}, {"key": "dc.rights.accesslevel", "value": "openAccess", "language": null, "element": "rights", "qualifier": "accesslevel", "schema": "dc"}, {"key": "dc.type.publication", "value": "masterThesis", "language": null, "element": "type", "qualifier": "publication", "schema": "dc"}, {"key": "dc.subject.oppiainekoodi", "value": "4041", "language": "", "element": "subject", "qualifier": "oppiainekoodi", "schema": "dc"}, {"key": "dc.subject.yso", "value": "neuroverkot", "language": null, "element": "subject", "qualifier": "yso", "schema": "dc"}, {"key": "dc.subject.yso", "value": "koneoppiminen", "language": null, "element": "subject", "qualifier": "yso", "schema": "dc"}, {"key": "dc.subject.yso", "value": "numeeriset menetelm\u00e4t", "language": null, "element": "subject", "qualifier": "yso", "schema": "dc"}, {"key": "dc.subject.yso", "value": "virheanalyysi", "language": null, "element": "subject", "qualifier": "yso", "schema": "dc"}, {"key": "dc.subject.yso", "value": "numeerinen analyysi", "language": null, "element": "subject", "qualifier": "yso", "schema": "dc"}, {"key": "dc.subject.yso", "value": "matematiikka", "language": null, "element": "subject", "qualifier": "yso", "schema": "dc"}, {"key": "dc.subject.yso", "value": "neural networks (information technology)", "language": null, "element": "subject", "qualifier": "yso", "schema": "dc"}, {"key": "dc.subject.yso", "value": "machine learning", "language": null, "element": "subject", "qualifier": "yso", "schema": "dc"}, {"key": "dc.subject.yso", "value": "numerical methods", "language": null, "element": "subject", "qualifier": "yso", "schema": "dc"}, {"key": "dc.subject.yso", "value": "error analysis", "language": null, "element": "subject", "qualifier": "yso", "schema": "dc"}, {"key": "dc.subject.yso", "value": "numerical analysis", "language": null, "element": "subject", "qualifier": "yso", "schema": "dc"}, {"key": "dc.subject.yso", "value": "mathematics", "language": null, "element": "subject", "qualifier": "yso", "schema": "dc"}, {"key": "dc.format.content", "value": "fulltext", "language": null, "element": "format", "qualifier": "content", "schema": "dc"}, {"key": "dc.rights.url", "value": "https://rightsstatements.org/page/InC/1.0/", "language": null, "element": "rights", "qualifier": "url", "schema": "dc"}, {"key": "dc.type.okm", "value": "G2", "language": null, "element": "type", "qualifier": "okm", "schema": "dc"}]
id jyx.123456789_78718
language eng
last_indexed 2025-03-31T20:02:28Z
main_date 2021-01-01T00:00:00Z
main_date_str 2021
online_boolean 1
online_urls_str_mv {"url":"https:\/\/jyx.jyu.fi\/bitstreams\/cd34de4d-cf05-41ab-86d8-d31f5baddaa5\/download","text":"URN:NBN:fi:jyu-202111195729.pdf","source":"jyx","mediaType":"application\/pdf"}
publishDate 2021
record_format qdc
source_str_mv jyx
spellingShingle Halonen, Vilho Accuracy analysis of uncertain variational problems with analytical and machine learning methods Matematiikka Mathematics 4041 neuroverkot koneoppiminen numeeriset menetelmät virheanalyysi numeerinen analyysi matematiikka neural networks (information technology) machine learning numerical methods error analysis numerical analysis mathematics
title Accuracy analysis of uncertain variational problems with analytical and machine learning methods
title_full Accuracy analysis of uncertain variational problems with analytical and machine learning methods
title_fullStr Accuracy analysis of uncertain variational problems with analytical and machine learning methods Accuracy analysis of uncertain variational problems with analytical and machine learning methods
title_full_unstemmed Accuracy analysis of uncertain variational problems with analytical and machine learning methods Accuracy analysis of uncertain variational problems with analytical and machine learning methods
title_short Accuracy analysis of uncertain variational problems with analytical and machine learning methods
title_sort accuracy analysis of uncertain variational problems with analytical and machine learning methods
title_txtP Accuracy analysis of uncertain variational problems with analytical and machine learning methods
topic Matematiikka Mathematics 4041 neuroverkot koneoppiminen numeeriset menetelmät virheanalyysi numeerinen analyysi matematiikka neural networks (information technology) machine learning numerical methods error analysis numerical analysis mathematics
topic_facet 4041 Matematiikka Mathematics error analysis koneoppiminen machine learning matematiikka mathematics neural networks (information technology) neuroverkot numeerinen analyysi numeeriset menetelmät numerical analysis numerical methods virheanalyysi
url https://jyx.jyu.fi/handle/123456789/78718 http://www.urn.fi/URN:NBN:fi:jyu-202111195729
work_keys_str_mv AT halonenvilho accuracyanalysisofuncertainvariationalproblemswithanalyticalandmachinelearningmetho