Probabilistic Deep Learning using Random Sum-Product Networks

The need for consistent treatment of uncertainty has recently triggered increased interest in probabilistic deep learning methods. However, most current approaches have severe limitations when it comes to inference, since many of these models do not even permit to evaluate exact data likelihoods. Su...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Peharz, Robert, Vergari, Antonio, Stelzner, Karl, Molina, Alejandro, Trapp, Martin, Kersting, Kristian, Ghahramani, Zoubin
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 Peharz, Robert
Vergari, Antonio
Stelzner, Karl
Molina, Alejandro
Trapp, Martin
Kersting, Kristian
Ghahramani, Zoubin
description The need for consistent treatment of uncertainty has recently triggered increased interest in probabilistic deep learning methods. However, most current approaches have severe limitations when it comes to inference, since many of these models do not even permit to evaluate exact data likelihoods. Sum-product networks (SPNs), on the other hand, are an excellent architecture in that regard, as they allow to efficiently evaluate likelihoods, as well as arbitrary marginalization and conditioning tasks. Nevertheless, SPNs have not been fully explored as serious deep learning models, likely due to their special structural requirements, which complicate learning. In this paper, we make a drastic simplification and use random SPN structures which are trained in a "classical deep learning manner", i.e. employing automatic differentiation, SGD, and GPU support. The resulting models, called RAT-SPNs, yield prediction results comparable to deep neural networks, while still being interpretable as generative model and maintaining well-calibrated uncertainties. This property makes them highly robust under missing input features and enables them to naturally detect outliers and peculiar samples.
doi_str_mv 10.48550/arxiv.1806.01910
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_1806_01910</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>1806_01910</sourcerecordid><originalsourceid>FETCH-LOGICAL-a670-fc763155e2dfa87fe67629874fb01b08424f9724ba12f8a80999c75c99cede013</originalsourceid><addsrcrecordid>eNotj8tOwzAURL1hgQofwAr_QMK14-eCBSpPKQLUdh9dOzayaJLKSXj8PW3p5sxmNJpDyBWDUhgp4QbzT_oqmQFVArMMzsntex4curRN45Q8vQ9hR-uAuU_9B53HA1fYt0NH13NX7Mvt7Cf6GqbvIX-OF-Qs4nYMl6dckM3jw2b5XNRvTy_Lu7pApaGIXquKSRl4G9HoGJRW3BotogPmwAguotVcOGQ8GjRgrfVa-j1DG4BVC3L9P3v83-xy6jD_NgeP5uhR_QEHGkJW</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Probabilistic Deep Learning using Random Sum-Product Networks</title><source>arXiv.org</source><creator>Peharz, Robert ; Vergari, Antonio ; Stelzner, Karl ; Molina, Alejandro ; Trapp, Martin ; Kersting, Kristian ; Ghahramani, Zoubin</creator><creatorcontrib>Peharz, Robert ; Vergari, Antonio ; Stelzner, Karl ; Molina, Alejandro ; Trapp, Martin ; Kersting, Kristian ; Ghahramani, Zoubin</creatorcontrib><description>The need for consistent treatment of uncertainty has recently triggered increased interest in probabilistic deep learning methods. However, most current approaches have severe limitations when it comes to inference, since many of these models do not even permit to evaluate exact data likelihoods. Sum-product networks (SPNs), on the other hand, are an excellent architecture in that regard, as they allow to efficiently evaluate likelihoods, as well as arbitrary marginalization and conditioning tasks. Nevertheless, SPNs have not been fully explored as serious deep learning models, likely due to their special structural requirements, which complicate learning. In this paper, we make a drastic simplification and use random SPN structures which are trained in a "classical deep learning manner", i.e. employing automatic differentiation, SGD, and GPU support. The resulting models, called RAT-SPNs, yield prediction results comparable to deep neural networks, while still being interpretable as generative model and maintaining well-calibrated uncertainties. This property makes them highly robust under missing input features and enables them to naturally detect outliers and peculiar samples.</description><identifier>DOI: 10.48550/arxiv.1806.01910</identifier><language>eng</language><subject>Computer Science - Artificial Intelligence ; Computer Science - Learning ; Statistics - Machine Learning</subject><creationdate>2018-06</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>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,881</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/1806.01910$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.1806.01910$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Peharz, Robert</creatorcontrib><creatorcontrib>Vergari, Antonio</creatorcontrib><creatorcontrib>Stelzner, Karl</creatorcontrib><creatorcontrib>Molina, Alejandro</creatorcontrib><creatorcontrib>Trapp, Martin</creatorcontrib><creatorcontrib>Kersting, Kristian</creatorcontrib><creatorcontrib>Ghahramani, Zoubin</creatorcontrib><title>Probabilistic Deep Learning using Random Sum-Product Networks</title><description>The need for consistent treatment of uncertainty has recently triggered increased interest in probabilistic deep learning methods. However, most current approaches have severe limitations when it comes to inference, since many of these models do not even permit to evaluate exact data likelihoods. Sum-product networks (SPNs), on the other hand, are an excellent architecture in that regard, as they allow to efficiently evaluate likelihoods, as well as arbitrary marginalization and conditioning tasks. Nevertheless, SPNs have not been fully explored as serious deep learning models, likely due to their special structural requirements, which complicate learning. In this paper, we make a drastic simplification and use random SPN structures which are trained in a "classical deep learning manner", i.e. employing automatic differentiation, SGD, and GPU support. The resulting models, called RAT-SPNs, yield prediction results comparable to deep neural networks, while still being interpretable as generative model and maintaining well-calibrated uncertainties. This property makes them highly robust under missing input features and enables them to naturally detect outliers and peculiar samples.</description><subject>Computer Science - Artificial Intelligence</subject><subject>Computer Science - Learning</subject><subject>Statistics - Machine Learning</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2018</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj8tOwzAURL1hgQofwAr_QMK14-eCBSpPKQLUdh9dOzayaJLKSXj8PW3p5sxmNJpDyBWDUhgp4QbzT_oqmQFVArMMzsntex4curRN45Q8vQ9hR-uAuU_9B53HA1fYt0NH13NX7Mvt7Cf6GqbvIX-OF-Qs4nYMl6dckM3jw2b5XNRvTy_Lu7pApaGIXquKSRl4G9HoGJRW3BotogPmwAguotVcOGQ8GjRgrfVa-j1DG4BVC3L9P3v83-xy6jD_NgeP5uhR_QEHGkJW</recordid><startdate>20180605</startdate><enddate>20180605</enddate><creator>Peharz, Robert</creator><creator>Vergari, Antonio</creator><creator>Stelzner, Karl</creator><creator>Molina, Alejandro</creator><creator>Trapp, Martin</creator><creator>Kersting, Kristian</creator><creator>Ghahramani, Zoubin</creator><scope>AKY</scope><scope>EPD</scope><scope>GOX</scope></search><sort><creationdate>20180605</creationdate><title>Probabilistic Deep Learning using Random Sum-Product Networks</title><author>Peharz, Robert ; Vergari, Antonio ; Stelzner, Karl ; Molina, Alejandro ; Trapp, Martin ; Kersting, Kristian ; Ghahramani, Zoubin</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a670-fc763155e2dfa87fe67629874fb01b08424f9724ba12f8a80999c75c99cede013</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2018</creationdate><topic>Computer Science - Artificial Intelligence</topic><topic>Computer Science - Learning</topic><topic>Statistics - Machine Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Peharz, Robert</creatorcontrib><creatorcontrib>Vergari, Antonio</creatorcontrib><creatorcontrib>Stelzner, Karl</creatorcontrib><creatorcontrib>Molina, Alejandro</creatorcontrib><creatorcontrib>Trapp, Martin</creatorcontrib><creatorcontrib>Kersting, Kristian</creatorcontrib><creatorcontrib>Ghahramani, Zoubin</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv Statistics</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Peharz, Robert</au><au>Vergari, Antonio</au><au>Stelzner, Karl</au><au>Molina, Alejandro</au><au>Trapp, Martin</au><au>Kersting, Kristian</au><au>Ghahramani, Zoubin</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Probabilistic Deep Learning using Random Sum-Product Networks</atitle><date>2018-06-05</date><risdate>2018</risdate><abstract>The need for consistent treatment of uncertainty has recently triggered increased interest in probabilistic deep learning methods. However, most current approaches have severe limitations when it comes to inference, since many of these models do not even permit to evaluate exact data likelihoods. Sum-product networks (SPNs), on the other hand, are an excellent architecture in that regard, as they allow to efficiently evaluate likelihoods, as well as arbitrary marginalization and conditioning tasks. Nevertheless, SPNs have not been fully explored as serious deep learning models, likely due to their special structural requirements, which complicate learning. In this paper, we make a drastic simplification and use random SPN structures which are trained in a "classical deep learning manner", i.e. employing automatic differentiation, SGD, and GPU support. The resulting models, called RAT-SPNs, yield prediction results comparable to deep neural networks, while still being interpretable as generative model and maintaining well-calibrated uncertainties. This property makes them highly robust under missing input features and enables them to naturally detect outliers and peculiar samples.</abstract><doi>10.48550/arxiv.1806.01910</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.1806.01910
ispartof
issn
language eng
recordid cdi_arxiv_primary_1806_01910
source arXiv.org
subjects Computer Science - Artificial Intelligence
Computer Science - Learning
Statistics - Machine Learning
title Probabilistic Deep Learning using Random Sum-Product Networks
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-27T09%3A01%3A05IST&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=Probabilistic%20Deep%20Learning%20using%20Random%20Sum-Product%20Networks&rft.au=Peharz,%20Robert&rft.date=2018-06-05&rft_id=info:doi/10.48550/arxiv.1806.01910&rft_dat=%3Carxiv_GOX%3E1806_01910%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