Publikation: On the Road to Clarity : Exploring Explainable AI for World Models in a Driver Assistance System
Dateien
Datum
Autor:innen
Herausgeber:innen
ISSN der Zeitschrift
Electronic ISSN
ISBN
Bibliografische Daten
Verlag
Schriftenreihe
Auflagebezeichnung
DOI (zitierfähiger Link)
Internationale Patentnummer
Angaben zur Forschungsförderung
Projekt
Open Access-Veröffentlichung
Core Facility der Universität Konstanz
Titel in einer weiteren Sprache
Publikationstyp
Publikationsstatus
Erschienen in
Zusammenfassung
In Autonomous Driving (AD) transparency and safety are paramount, as mistakes are costly. However, neural networks used in AD systems are generally considered black boxes. As a countermeasure, we have methods of explainable AI (XAI), such as feature relevance estimation and dimensionality reduction. Coarse graining techniques can also help reduce dimensionality and find interpretable global patterns. A specific coarse graining method is Renormalization Groups from statistical physics. It has previously been applied to Restricted Boltzmann Machines (RBMs) to interpret unsupervised learning. We refine this technique by building a transparent backbone model for convolutional variational autoencoders (VAE) that allows mapping latent values to input features and has performance comparable to trained black box VAEs. Moreover, we propose a custom feature map visualization technique to analyze the internal convolutional layers in the VAE to explain internal causes of poor reconstruction that may lead to dangerous traffic scenarios in AD applications. In a second key contribution, we propose explanation and evaluation techniques for the internal dynamics and feature relevance of prediction networks. We test a long short-term memory (LSTM) network in the computer vision domain to evaluate the predictability and in future applications potentially safety of prediction models. We showcase our methods by analyzing a VAE-LSTM world model that predicts pedestrian perception in an urban traffic situation.
Zusammenfassung in einer weiteren Sprache
Fachgebiet (DDC)
Schlagwörter
Konferenz
Rezension
Zitieren
ISO 690
ROSHDI, Mohamed, Julian PETZOLD, Mostafa WAHBY, Hussein EBRAHIM, Mladen BEREKOVIC, Heiko HAMANN, 2024. On the Road to Clarity : Exploring Explainable AI for World Models in a Driver Assistance System. CAI 2024 : IEEE Conference on Artificial Intelligence. Marina Bay Sands, Singapore, 25. Juni 2024 - 27. Juni 2024. In: NERI, Ferrante, Hrsg., Guansong PANG, Hrsg., Mengmi ZHANG, Hrsg.. 2024 IEEE Conference on Artificial Intelligence : CAI 2024 : 25-27 June 2024, Marina Bay Sands, Singapore, proceedings. Piscataway, NJ: IEEE, 2024, S. 1032-1039. ISBN 979-8-3503-5410-2. Verfügbar unter: doi: 10.1109/cai59869.2024.00187BibTex
@inproceedings{Roshdi2024-06-25Clari-71403, year={2024}, doi={10.1109/cai59869.2024.00187}, title={On the Road to Clarity : Exploring Explainable AI for World Models in a Driver Assistance System}, isbn={979-8-3503-5410-2}, publisher={IEEE}, address={Piscataway, NJ}, booktitle={2024 IEEE Conference on Artificial Intelligence : CAI 2024 : 25-27 June 2024, Marina Bay Sands, Singapore, proceedings}, pages={1032--1039}, editor={Neri, Ferrante and Pang, Guansong and Zhang, Mengmi}, author={Roshdi, Mohamed and Petzold, Julian and Wahby, Mostafa and Ebrahim, Hussein and Berekovic, Mladen and Hamann, Heiko} }
RDF
<rdf:RDF xmlns:dcterms="http://purl.org/dc/terms/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:bibo="http://purl.org/ontology/bibo/" xmlns:dspace="http://digital-repositories.org/ontologies/dspace/0.1.0#" xmlns:foaf="http://xmlns.com/foaf/0.1/" xmlns:void="http://rdfs.org/ns/void#" xmlns:xsd="http://www.w3.org/2001/XMLSchema#" > <rdf:Description rdf:about="https://kops.uni-konstanz.de/server/rdf/resource/123456789/71403"> <dcterms:isPartOf rdf:resource="https://kops.uni-konstanz.de/server/rdf/resource/123456789/36"/> <dspace:isPartOfCollection rdf:resource="https://kops.uni-konstanz.de/server/rdf/resource/123456789/36"/> <foaf:homepage rdf:resource="http://localhost:8080/"/> <dcterms:title>On the Road to Clarity : Exploring Explainable AI for World Models in a Driver Assistance System</dcterms:title> <dc:creator>Hamann, Heiko</dc:creator> <dc:language>eng</dc:language> <dc:creator>Wahby, Mostafa</dc:creator> <dc:contributor>Wahby, Mostafa</dc:contributor> <dc:creator>Petzold, Julian</dc:creator> <dc:creator>Ebrahim, Hussein</dc:creator> <dc:contributor>Petzold, Julian</dc:contributor> <dcterms:abstract>In Autonomous Driving (AD) transparency and safety are paramount, as mistakes are costly. However, neural networks used in AD systems are generally considered black boxes. As a countermeasure, we have methods of explainable AI (XAI), such as feature relevance estimation and dimensionality reduction. Coarse graining techniques can also help reduce dimensionality and find interpretable global patterns. A specific coarse graining method is Renormalization Groups from statistical physics. It has previously been applied to Restricted Boltzmann Machines (RBMs) to interpret unsupervised learning. We refine this technique by building a transparent backbone model for convolutional variational autoencoders (VAE) that allows mapping latent values to input features and has performance comparable to trained black box VAEs. Moreover, we propose a custom feature map visualization technique to analyze the internal convolutional layers in the VAE to explain internal causes of poor reconstruction that may lead to dangerous traffic scenarios in AD applications. In a second key contribution, we propose explanation and evaluation techniques for the internal dynamics and feature relevance of prediction networks. We test a long short-term memory (LSTM) network in the computer vision domain to evaluate the predictability and in future applications potentially safety of prediction models. We showcase our methods by analyzing a VAE-LSTM world model that predicts pedestrian perception in an urban traffic situation.</dcterms:abstract> <dc:contributor>Berekovic, Mladen</dc:contributor> <void:sparqlEndpoint rdf:resource="http://localhost/fuseki/dspace/sparql"/> <bibo:uri rdf:resource="https://kops.uni-konstanz.de/handle/123456789/71403"/> <dc:contributor>Roshdi, Mohamed</dc:contributor> <dc:creator>Roshdi, Mohamed</dc:creator> <dc:contributor>Ebrahim, Hussein</dc:contributor> <dcterms:issued>2024-06-25</dcterms:issued> <dc:creator>Berekovic, Mladen</dc:creator> <dc:contributor>Hamann, Heiko</dc:contributor> <dc:date rdf:datatype="http://www.w3.org/2001/XMLSchema#dateTime">2024-11-22T10:08:57Z</dc:date> <dcterms:available rdf:datatype="http://www.w3.org/2001/XMLSchema#dateTime">2024-11-22T10:08:57Z</dcterms:available> </rdf:Description> </rdf:RDF>