PredDiff: Explanations and Interactions from Conditional Expectations

PredDiff: Explanations and Interactions from Conditional Expectations

PredDiff is a model-agnostic, local attribution method that is firmly rooted in probability theory. Its simple intuition is to measure prediction changes while marginalizing features. In this work, we clarify properties of PredDiff and its close connection to Shapley values. We stress important diff...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:arXiv.org 2022-09
Hauptverfasser: Blücher, Stefan, Vielhaben, Johanna, Strodthoff, Nils
Format: Artikel
Sprache:eng
Schlagworte:
Computer Science - Artificial Intelligence
Computer Science - Learning
Mathematical models
Probability theory
Regression analysis
Statistical analysis
Statistics - Machine Learning
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title arXiv.org
container_volume
creator Blücher, Stefan
Vielhaben, Johanna
Strodthoff, Nils
description PredDiff is a model-agnostic, local attribution method that is firmly rooted in probability theory. Its simple intuition is to measure prediction changes while marginalizing features. In this work, we clarify properties of PredDiff and its close connection to Shapley values. We stress important differences between classification and regression, which require a specific treatment within both formalisms. We extend PredDiff by introducing a new, well-founded measure for interaction effects between arbitrary feature subsets. The study of interaction effects represents an inevitable step towards a comprehensive understanding of black-box models and is particularly important for science applications. Equipped with our novel interaction measure, PredDiff is a promising model-agnostic approach for obtaining reliable, numerically inexpensive and theoretically sound attributions.
doi_str_mv 10.48550/arxiv.2102.13519
format Article
fullrecord <record><control><sourceid>proquest_arxiv</sourceid><recordid>TN_cdi_arxiv_primary_2102_13519</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2494716220</sourcerecordid><originalsourceid>FETCH-LOGICAL-a520-7e60f52e13675db0631d4ea27896cd03ecf89b70cab6daeb2dbfe09e0ff9d92b3</originalsourceid><addsrcrecordid>eNotj1FLwzAUhYMgOOZ-gE8WfG69uWmSxjepdQ4G-rD3kjQJdHRtTTqZ_95t9elyLh-H8xHyQCHLC87hWYdT-5MhBcwo41TdkAUyRtMiR7wjqxj3AIBCIudsQaqv4Oxb6_1LUp3GTvd6aoc-Jrq3yaafXNDN_PBhOCTl0Nv2knV3wV0zzfg9ufW6i271f5dk917tyo90-7nelK_bVHOEVDoBnqOjTEhuDQhGbe40ykKJxgJzjS-UkdBoI6x2Bq3xDpQD75VVaNiSPM61V8d6DO1Bh9_64lpfXc_E00yMYfg-ujjV--EYznNjjbnKJRWIwP4Adq1X8g</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2494716220</pqid></control><display><type>article</type><title>PredDiff: Explanations and Interactions from Conditional Expectations</title><source>arXiv.org</source><source>Free E- Journals</source><creator>Blücher, Stefan ; Vielhaben, Johanna ; Strodthoff, Nils</creator><creatorcontrib>Blücher, Stefan ; Vielhaben, Johanna ; Strodthoff, Nils</creatorcontrib><description>PredDiff is a model-agnostic, local attribution method that is firmly rooted in probability theory. Its simple intuition is to measure prediction changes while marginalizing features. In this work, we clarify properties of PredDiff and its close connection to Shapley values. We stress important differences between classification and regression, which require a specific treatment within both formalisms. We extend PredDiff by introducing a new, well-founded measure for interaction effects between arbitrary feature subsets. The study of interaction effects represents an inevitable step towards a comprehensive understanding of black-box models and is particularly important for science applications. Equipped with our novel interaction measure, PredDiff is a promising model-agnostic approach for obtaining reliable, numerically inexpensive and theoretically sound attributions.</description><identifier>EISSN: 2331-8422</identifier><identifier>DOI: 10.48550/arxiv.2102.13519</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Computer Science - Artificial Intelligence ; Computer Science - Learning ; Mathematical models ; Probability theory ; Regression analysis ; Statistical analysis ; Statistics - Machine Learning</subject><ispartof>arXiv.org, 2022-09</ispartof><rights>2022. This work is published under http://creativecommons.org/licenses/by-nc-nd/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><rights>http://creativecommons.org/licenses/by-nc-nd/4.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,780,881,27904</link.rule.ids><backlink>$$Uhttps://doi.org/10.1016/j.artint.2022.103774$$DView published paper (Access to full text may be restricted)$$Hfree_for_read</backlink><backlink>$$Uhttps://doi.org/10.48550/arXiv.2102.13519$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Blücher, Stefan</creatorcontrib><creatorcontrib>Vielhaben, Johanna</creatorcontrib><creatorcontrib>Strodthoff, Nils</creatorcontrib><title>PredDiff: Explanations and Interactions from Conditional Expectations</title><title>arXiv.org</title><description>PredDiff is a model-agnostic, local attribution method that is firmly rooted in probability theory. Its simple intuition is to measure prediction changes while marginalizing features. In this work, we clarify properties of PredDiff and its close connection to Shapley values. We stress important differences between classification and regression, which require a specific treatment within both formalisms. We extend PredDiff by introducing a new, well-founded measure for interaction effects between arbitrary feature subsets. The study of interaction effects represents an inevitable step towards a comprehensive understanding of black-box models and is particularly important for science applications. Equipped with our novel interaction measure, PredDiff is a promising model-agnostic approach for obtaining reliable, numerically inexpensive and theoretically sound attributions.</description><subject>Computer Science - Artificial Intelligence</subject><subject>Computer Science - Learning</subject><subject>Mathematical models</subject><subject>Probability theory</subject><subject>Regression analysis</subject><subject>Statistical analysis</subject><subject>Statistics - Machine Learning</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2022</creationdate><recordtype>article</recordtype><sourceid>BENPR</sourceid><sourceid>GOX</sourceid><recordid>eNotj1FLwzAUhYMgOOZ-gE8WfG69uWmSxjepdQ4G-rD3kjQJdHRtTTqZ_95t9elyLh-H8xHyQCHLC87hWYdT-5MhBcwo41TdkAUyRtMiR7wjqxj3AIBCIudsQaqv4Oxb6_1LUp3GTvd6aoc-Jrq3yaafXNDN_PBhOCTl0Nv2knV3wV0zzfg9ufW6i271f5dk917tyo90-7nelK_bVHOEVDoBnqOjTEhuDQhGbe40ykKJxgJzjS-UkdBoI6x2Bq3xDpQD75VVaNiSPM61V8d6DO1Bh9_64lpfXc_E00yMYfg-ujjV--EYznNjjbnKJRWIwP4Adq1X8g</recordid><startdate>20220908</startdate><enddate>20220908</enddate><creator>Blücher, Stefan</creator><creator>Vielhaben, Johanna</creator><creator>Strodthoff, Nils</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope><scope>AKY</scope><scope>EPD</scope><scope>GOX</scope></search><sort><creationdate>20220908</creationdate><title>PredDiff: Explanations and Interactions from Conditional Expectations</title><author>Blücher, Stefan ; Vielhaben, Johanna ; Strodthoff, Nils</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a520-7e60f52e13675db0631d4ea27896cd03ecf89b70cab6daeb2dbfe09e0ff9d92b3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2022</creationdate><topic>Computer Science - Artificial Intelligence</topic><topic>Computer Science - Learning</topic><topic>Mathematical models</topic><topic>Probability theory</topic><topic>Regression analysis</topic><topic>Statistical analysis</topic><topic>Statistics - Machine Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Blücher, Stefan</creatorcontrib><creatorcontrib>Vielhaben, Johanna</creatorcontrib><creatorcontrib>Strodthoff, Nils</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science &amp; Engineering Collection</collection><collection>ProQuest Central (Alumni Edition)</collection><collection>ProQuest Central UK/Ireland</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central Korea</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Publicly Available Content Database</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 China</collection><collection>Engineering Collection</collection><collection>arXiv Computer Science</collection><collection>arXiv Statistics</collection><collection>arXiv.org</collection><jtitle>arXiv.org</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Blücher, Stefan</au><au>Vielhaben, Johanna</au><au>Strodthoff, Nils</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>PredDiff: Explanations and Interactions from Conditional Expectations</atitle><jtitle>arXiv.org</jtitle><date>2022-09-08</date><risdate>2022</risdate><eissn>2331-8422</eissn><abstract>PredDiff is a model-agnostic, local attribution method that is firmly rooted in probability theory. Its simple intuition is to measure prediction changes while marginalizing features. In this work, we clarify properties of PredDiff and its close connection to Shapley values. We stress important differences between classification and regression, which require a specific treatment within both formalisms. We extend PredDiff by introducing a new, well-founded measure for interaction effects between arbitrary feature subsets. The study of interaction effects represents an inevitable step towards a comprehensive understanding of black-box models and is particularly important for science applications. Equipped with our novel interaction measure, PredDiff is a promising model-agnostic approach for obtaining reliable, numerically inexpensive and theoretically sound attributions.</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><doi>10.48550/arxiv.2102.13519</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier EISSN: 2331-8422
ispartof arXiv.org, 2022-09
issn 2331-8422
language eng
recordid cdi_arxiv_primary_2102_13519
source arXiv.org; Free E- Journals
subjects Computer Science - Artificial Intelligence
Computer Science - Learning
Mathematical models
Probability theory
Regression analysis
Statistical analysis
Statistics - Machine Learning
title PredDiff: Explanations and Interactions from Conditional Expectations
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-28T05%3A13%3A14IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_arxiv&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=PredDiff:%20Explanations%20and%20Interactions%20from%20Conditional%20Expectations&rft.jtitle=arXiv.org&rft.au=Bl%C3%BCcher,%20Stefan&rft.date=2022-09-08&rft.eissn=2331-8422&rft_id=info:doi/10.48550/arxiv.2102.13519&rft_dat=%3Cproquest_arxiv%3E2494716220%3C/proquest_arxiv%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2494716220&rft_id=info:pmid/&rfr_iscdi=true