Keep the Faith: Faithful Explanations in Convolutional Neural Networks for Case-Based Reasoning

Explaining predictions of black-box neural networks is crucial when applied to decision-critical tasks. Thus, attribution maps are commonly used to identify important image regions, despite prior work showing that humans prefer explanations based on similar examples. To this end, ProtoPNet learns a...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Wolf, Tom Nuno, Bongratz, Fabian, Rickmann, Anne-Marie, Pölsterl, Sebastian, Wachinger, Christian
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 Wolf, Tom Nuno
Bongratz, Fabian
Rickmann, Anne-Marie
Pölsterl, Sebastian
Wachinger, Christian
description Explaining predictions of black-box neural networks is crucial when applied to decision-critical tasks. Thus, attribution maps are commonly used to identify important image regions, despite prior work showing that humans prefer explanations based on similar examples. To this end, ProtoPNet learns a set of class-representative feature vectors (prototypes) for case-based reasoning. During inference, similarities of latent features to prototypes are linearly classified to form predictions and attribution maps are provided to explain the similarity. In this work, we evaluate whether architectures for case-based reasoning fulfill established axioms required for faithful explanations using the example of ProtoPNet. We show that such architectures allow the extraction of faithful explanations. However, we prove that the attribution maps used to explain the similarities violate the axioms. We propose a new procedure to extract explanations for trained ProtoPNets, named ProtoPFaith. Conceptually, these explanations are Shapley values, calculated on the similarity scores of each prototype. They allow to faithfully answer which prototypes are present in an unseen image and quantify each pixel's contribution to that presence, thereby complying with all axioms. The theoretical violations of ProtoPNet manifest in our experiments on three datasets (CUB-200-2011, Stanford Dogs, RSNA) and five architectures (ConvNet, ResNet, ResNet50, WideResNet50, ResNeXt50). Our experiments show a qualitative difference between the explanations given by ProtoPNet and ProtoPFaith. Additionally, we quantify the explanations with the Area Over the Perturbation Curve, on which ProtoPFaith outperforms ProtoPNet on all experiments by a factor $>10^3$.
doi_str_mv 10.48550/arxiv.2312.09783
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2312_09783</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2312_09783</sourcerecordid><originalsourceid>FETCH-LOGICAL-a673-858f1e463d380b058df5b621cd05d78ff26a3a0ec6352493f8660e117aa091263</originalsourceid><addsrcrecordid>eNotj71OwzAYRb0woMIDMOEXSPBP7DhsELWAWoFUdY--1nZrYezITkp5e2jKco_ucqSD0B0lZaWEIA-QTu5YMk5ZSZpa8WvULY3p8XAweAFuODxeYEeP56feQ4DBxZCxC7iN4Rj9eP7g8bsZ04ThO6bPjG1MuIVsiue_0XhtIMfgwv4GXVnw2dz-c4Y2i_mmfS1WHy9v7dOqAFnzQgllqakk11yRLRFKW7GVjO40EbpW1jIJHIjZSS5Y1XCrpCSG0hqANJRJPkP3F-1U2PXJfUH66c6l3VTKfwEnbk4d</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Keep the Faith: Faithful Explanations in Convolutional Neural Networks for Case-Based Reasoning</title><source>arXiv.org</source><creator>Wolf, Tom Nuno ; Bongratz, Fabian ; Rickmann, Anne-Marie ; Pölsterl, Sebastian ; Wachinger, Christian</creator><creatorcontrib>Wolf, Tom Nuno ; Bongratz, Fabian ; Rickmann, Anne-Marie ; Pölsterl, Sebastian ; Wachinger, Christian</creatorcontrib><description>Explaining predictions of black-box neural networks is crucial when applied to decision-critical tasks. Thus, attribution maps are commonly used to identify important image regions, despite prior work showing that humans prefer explanations based on similar examples. To this end, ProtoPNet learns a set of class-representative feature vectors (prototypes) for case-based reasoning. During inference, similarities of latent features to prototypes are linearly classified to form predictions and attribution maps are provided to explain the similarity. In this work, we evaluate whether architectures for case-based reasoning fulfill established axioms required for faithful explanations using the example of ProtoPNet. We show that such architectures allow the extraction of faithful explanations. However, we prove that the attribution maps used to explain the similarities violate the axioms. We propose a new procedure to extract explanations for trained ProtoPNets, named ProtoPFaith. Conceptually, these explanations are Shapley values, calculated on the similarity scores of each prototype. They allow to faithfully answer which prototypes are present in an unseen image and quantify each pixel's contribution to that presence, thereby complying with all axioms. The theoretical violations of ProtoPNet manifest in our experiments on three datasets (CUB-200-2011, Stanford Dogs, RSNA) and five architectures (ConvNet, ResNet, ResNet50, WideResNet50, ResNeXt50). Our experiments show a qualitative difference between the explanations given by ProtoPNet and ProtoPFaith. Additionally, we quantify the explanations with the Area Over the Perturbation Curve, on which ProtoPFaith outperforms ProtoPNet on all experiments by a factor $&gt;10^3$.</description><identifier>DOI: 10.48550/arxiv.2312.09783</identifier><language>eng</language><subject>Computer Science - Computer Vision and Pattern Recognition ; Computer Science - Learning</subject><creationdate>2023-12</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>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,778,883</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2312.09783$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2312.09783$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Wolf, Tom Nuno</creatorcontrib><creatorcontrib>Bongratz, Fabian</creatorcontrib><creatorcontrib>Rickmann, Anne-Marie</creatorcontrib><creatorcontrib>Pölsterl, Sebastian</creatorcontrib><creatorcontrib>Wachinger, Christian</creatorcontrib><title>Keep the Faith: Faithful Explanations in Convolutional Neural Networks for Case-Based Reasoning</title><description>Explaining predictions of black-box neural networks is crucial when applied to decision-critical tasks. Thus, attribution maps are commonly used to identify important image regions, despite prior work showing that humans prefer explanations based on similar examples. To this end, ProtoPNet learns a set of class-representative feature vectors (prototypes) for case-based reasoning. During inference, similarities of latent features to prototypes are linearly classified to form predictions and attribution maps are provided to explain the similarity. In this work, we evaluate whether architectures for case-based reasoning fulfill established axioms required for faithful explanations using the example of ProtoPNet. We show that such architectures allow the extraction of faithful explanations. However, we prove that the attribution maps used to explain the similarities violate the axioms. We propose a new procedure to extract explanations for trained ProtoPNets, named ProtoPFaith. Conceptually, these explanations are Shapley values, calculated on the similarity scores of each prototype. They allow to faithfully answer which prototypes are present in an unseen image and quantify each pixel's contribution to that presence, thereby complying with all axioms. The theoretical violations of ProtoPNet manifest in our experiments on three datasets (CUB-200-2011, Stanford Dogs, RSNA) and five architectures (ConvNet, ResNet, ResNet50, WideResNet50, ResNeXt50). Our experiments show a qualitative difference between the explanations given by ProtoPNet and ProtoPFaith. Additionally, we quantify the explanations with the Area Over the Perturbation Curve, on which ProtoPFaith outperforms ProtoPNet on all experiments by a factor $&gt;10^3$.</description><subject>Computer Science - Computer Vision and Pattern Recognition</subject><subject>Computer Science - Learning</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj71OwzAYRb0woMIDMOEXSPBP7DhsELWAWoFUdY--1nZrYezITkp5e2jKco_ucqSD0B0lZaWEIA-QTu5YMk5ZSZpa8WvULY3p8XAweAFuODxeYEeP56feQ4DBxZCxC7iN4Rj9eP7g8bsZ04ThO6bPjG1MuIVsiue_0XhtIMfgwv4GXVnw2dz-c4Y2i_mmfS1WHy9v7dOqAFnzQgllqakk11yRLRFKW7GVjO40EbpW1jIJHIjZSS5Y1XCrpCSG0hqANJRJPkP3F-1U2PXJfUH66c6l3VTKfwEnbk4d</recordid><startdate>20231215</startdate><enddate>20231215</enddate><creator>Wolf, Tom Nuno</creator><creator>Bongratz, Fabian</creator><creator>Rickmann, Anne-Marie</creator><creator>Pölsterl, Sebastian</creator><creator>Wachinger, Christian</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20231215</creationdate><title>Keep the Faith: Faithful Explanations in Convolutional Neural Networks for Case-Based Reasoning</title><author>Wolf, Tom Nuno ; Bongratz, Fabian ; Rickmann, Anne-Marie ; Pölsterl, Sebastian ; Wachinger, Christian</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a673-858f1e463d380b058df5b621cd05d78ff26a3a0ec6352493f8660e117aa091263</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Computer Science - Computer Vision and Pattern Recognition</topic><topic>Computer Science - Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Wolf, Tom Nuno</creatorcontrib><creatorcontrib>Bongratz, Fabian</creatorcontrib><creatorcontrib>Rickmann, Anne-Marie</creatorcontrib><creatorcontrib>Pölsterl, Sebastian</creatorcontrib><creatorcontrib>Wachinger, Christian</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Wolf, Tom Nuno</au><au>Bongratz, Fabian</au><au>Rickmann, Anne-Marie</au><au>Pölsterl, Sebastian</au><au>Wachinger, Christian</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Keep the Faith: Faithful Explanations in Convolutional Neural Networks for Case-Based Reasoning</atitle><date>2023-12-15</date><risdate>2023</risdate><abstract>Explaining predictions of black-box neural networks is crucial when applied to decision-critical tasks. Thus, attribution maps are commonly used to identify important image regions, despite prior work showing that humans prefer explanations based on similar examples. To this end, ProtoPNet learns a set of class-representative feature vectors (prototypes) for case-based reasoning. During inference, similarities of latent features to prototypes are linearly classified to form predictions and attribution maps are provided to explain the similarity. In this work, we evaluate whether architectures for case-based reasoning fulfill established axioms required for faithful explanations using the example of ProtoPNet. We show that such architectures allow the extraction of faithful explanations. However, we prove that the attribution maps used to explain the similarities violate the axioms. We propose a new procedure to extract explanations for trained ProtoPNets, named ProtoPFaith. Conceptually, these explanations are Shapley values, calculated on the similarity scores of each prototype. They allow to faithfully answer which prototypes are present in an unseen image and quantify each pixel's contribution to that presence, thereby complying with all axioms. The theoretical violations of ProtoPNet manifest in our experiments on three datasets (CUB-200-2011, Stanford Dogs, RSNA) and five architectures (ConvNet, ResNet, ResNet50, WideResNet50, ResNeXt50). Our experiments show a qualitative difference between the explanations given by ProtoPNet and ProtoPFaith. Additionally, we quantify the explanations with the Area Over the Perturbation Curve, on which ProtoPFaith outperforms ProtoPNet on all experiments by a factor $&gt;10^3$.</abstract><doi>10.48550/arxiv.2312.09783</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2312.09783
ispartof
issn
language eng
recordid cdi_arxiv_primary_2312_09783
source arXiv.org
subjects Computer Science - Computer Vision and Pattern Recognition
Computer Science - Learning
title Keep the Faith: Faithful Explanations in Convolutional Neural Networks for Case-Based Reasoning
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-15T11%3A24%3A53IST&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=Keep%20the%20Faith:%20Faithful%20Explanations%20in%20Convolutional%20Neural%20Networks%20for%20Case-Based%20Reasoning&rft.au=Wolf,%20Tom%20Nuno&rft.date=2023-12-15&rft_id=info:doi/10.48550/arxiv.2312.09783&rft_dat=%3Carxiv_GOX%3E2312_09783%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