東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Neural Methods for Link Prediction i...

Kochsiek, Adrian,

Linked to FindBook

Google Book

Amazon

博客來

Neural Methods for Link Prediction in Knowledge Graphs /

Record Type:	Electronic resources : Monograph/item
Title/Author:	Neural Methods for Link Prediction in Knowledge Graphs // Adrian Kochsiek.
Author:	Kochsiek, Adrian,
Description:	1 electronic resource (169 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 86-10, Section: A.
Contained By:	Dissertations Abstracts International86-10A.
Subject:	Graphs. -
Online resource:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31963743
ISBN:	9798311954990

Neural Methods for Link Prediction in Knowledge Graphs /
Kochsiek, Adrian,

Neural Methods for Link Prediction in Knowledge Graphs /Adrian Kochsiek. - 1 electronic resource (169 pages)

Source: Dissertations Abstracts International, Volume: 86-10, Section: A.

Integrating neural models with multi-relational data poses a significant challenge, mainly due to the complexity and highly interconnected nature of such data, which is prevalent in domains such as social networks, knowledge graphs, and biomedical databases. In this context, this thesis explores three key challenges: effectively capturing multi-relational data for efficient reasoning and integration into downstream applications, integrating structural and textual information, and managing the evolving nature of multi-relational graphs. Knowledge graph embeddings (KGE) offer a promising approach, learning low-dimensional representations for entities and relations to enhance tasks such as recommendation, question answering, and visual relationship detection. However, scalability and quality assurance in KGE models remain issues, especially with large-scale graphs. Recent frameworks address scalability through parallelization techniques, but their impact on model quality requires further study. While extensive hyperparameter optimization can considerably improve the resulting embedding quality, current optimization methods do not scale to large-scale graphs. Moreover, integrating textual information with structural data in a simple, scalable, and versatile manner remains an open problem.Addressing these challenges, this thesis makes several key contributions. First, it evaluates and improves the efficiency and effectiveness of parallel training techniques for KGE models. By re-implementing and extensively testing various parallelization methods, the study identifies optimal techniques for large-scale KGE training, demonstrating significant speedups while maintaining model quality. Second, it proposes GraSH, a scalable hyperparameter optimization algorithm based on successive halving, achieving state-of-the-art results on large-scale knowledge graphs with minimal search budgets. Third, the thesis introduces KGT5-context, a novel approach that integrates structural and textual information by posing link prediction as a sequence-to-sequence task, achieving or surpassing state-of-the-art results in link prediction and downstream tasks. Finally, it presents the Wikidata5M-SI benchmark for evaluating model performance on dynamic graphs with emerging entities in a realistic, large-scale setting.

English

ISBN: 9798311954990Subjects--Topical Terms:

3685259
Graphs.

Neural Methods for Link Prediction in Knowledge Graphs /
LDR:06257nmm a22003733i 4500 001 2433297
005 20250825104043.5
006 m o d
007 cr|nu||||||||
008 261011s2025 miu||||||m |||||||eng d
020 $a 9798311954990
035 $a (MiAaPQD)AAI31963743
035 $a (MiAaPQD)Mannheim68884
035 $a AAI31963743
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Kochsiek, Adrian, $e author. $3 3822848
245 1 0 $a Neural Methods for Link Prediction in Knowledge Graphs / $c Adrian Kochsiek.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2025
300 $a 1 electronic resource (169 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 86-10, Section: A.
500 $a Advisors: Gemulla, Rainer.
502 $b Ph.D. $c Universitaet Mannheim (Germany) $d 2025.
520 $a Integrating neural models with multi-relational data poses a significant challenge, mainly due to the complexity and highly interconnected nature of such data, which is prevalent in domains such as social networks, knowledge graphs, and biomedical databases. In this context, this thesis explores three key challenges: effectively capturing multi-relational data for efficient reasoning and integration into downstream applications, integrating structural and textual information, and managing the evolving nature of multi-relational graphs. Knowledge graph embeddings (KGE) offer a promising approach, learning low-dimensional representations for entities and relations to enhance tasks such as recommendation, question answering, and visual relationship detection. However, scalability and quality assurance in KGE models remain issues, especially with large-scale graphs. Recent frameworks address scalability through parallelization techniques, but their impact on model quality requires further study. While extensive hyperparameter optimization can considerably improve the resulting embedding quality, current optimization methods do not scale to large-scale graphs. Moreover, integrating textual information with structural data in a simple, scalable, and versatile manner remains an open problem.Addressing these challenges, this thesis makes several key contributions. First, it evaluates and improves the efficiency and effectiveness of parallel training techniques for KGE models. By re-implementing and extensively testing various parallelization methods, the study identifies optimal techniques for large-scale KGE training, demonstrating significant speedups while maintaining model quality. Second, it proposes GraSH, a scalable hyperparameter optimization algorithm based on successive halving, achieving state-of-the-art results on large-scale knowledge graphs with minimal search budgets. Third, the thesis introduces KGT5-context, a novel approach that integrates structural and textual information by posing link prediction as a sequence-to-sequence task, achieving or surpassing state-of-the-art results in link prediction and downstream tasks. Finally, it presents the Wikidata5M-SI benchmark for evaluating model performance on dynamic graphs with emerging entities in a realistic, large-scale setting.
520 $a Die Integration von neuronalen Modellen mit multi-relationalen Daten stellt eine bedeutende Herausforderung dar, aufgrund der Komplexitat der verbundenen Daten, wie sie in Bereichen wie sozialen Netzwerken, Wissensgraphen und biomedizinischen Datenbanken vorkommen. In diesem Kontext untersucht diese Arbeit drei zentrale Herausforderungen: (i) das effektive Erfassen von multi-relationalen Daten fur effiziente Inferenz und die Integration in nachgelagerte Anwendungen, (ii) die Integration von strukturellen und textuellen Informationen sowie (iii) die Handhabung der dynamischen Eigenschaft von multi-relationalen Graphen. Knowledge Graph Embeddings (KGE) bieten einen vielversprechenden Ansatz, indem sie niedrigdimensionale Reprasentationen fur Entitaten und Relationen lernen, um Aufgaben wie Empfehlung, Fragebeantwortung und visuelle Beziehungserkennung zu verbessern. Allerdings bleiben Skalierbarkeit und Qualitatssicherung bei KGE-Modellen Herausforderungen. Neue Frameworks adressieren die Skalierbarkeit durch Parallelisierungstechniken, aber deren Auswirkungen auf die Modellqualitat bedurfen weiterer Untersuchung. Wahrend umfangreiche Hyperparameter-Optimierung die Qualitat der resultierenden Embeddings erheblich verbessern kann, skalieren aktuelle Optimierungsmethoden nicht auf grose Graphen. Zudem bleibt die Integration von textuellen Informationen mit strukturellen Daten auf einfache, skalierbare und flexible Weise ein ungelostes Problem.Zur Bewaltigung dieser Herausforderungen leistet diese Arbeit mehrere wesentliche Beitrage. Erstens bewertet und verbessert sie die Effizienz und Effektivitat paralleler Trainingstechniken fur KGE-Modelle. Durch die Neuimplementierung und umfangreiche Prufung verschiedener Parallelisierungsmethoden identifiziert die Studie optimale Techniken fur das Training von KGE Modellen auf grosen Graphen und zeigt eine signifikante Reduzierung der Trainingszeit bei gleichzeitiger Erhaltung der Modellqualitat. Zweitens wird GraSH eingefuhrt, ein skalierbarer Hyperparameter-Optimierungsalgorithmus basierend auf "Successive Halving", der mit minimalen Suchbudget Ergebnisse auf dem Stand der Technik auf grosen Wissensgraphen erzielt. Drittens fuhrt die Arbeit KGT5-Kontext ein, einen neuartigen Ansatz, der strukturelle und textuelle Informationen integriert, indem er die Vorhersage von neuen Verbindungen im Graphen als Sequenz-zu-Sequenz-Aufgabe darstellt und dabei den derzeitigen Stand der Technik in der Linkvorhersage und darauf aufbauenden Aufgaben erreicht oder ubertrifft. Schlieslich wird die Wikidata5M-SI-Benchmark vorgestellt, um die Modellleistung auf grosen, dynamischen Graphen mit neuen Entitaten in einem realistischen Setting zu evaluieren.
546 $a English
590 $a School code: 1124
650 4 $a Graphs. $3 3685259
650 4 $a Ontology. $3 530874
650 4 $a Recommender systems. $3 3562220
650 4 $a Benchmarks. $3 3548256
650 4 $a Linguistics. $3 524476
650 4 $a Visualization. $3 586179
650 4 $a Knowledge representation. $3 3689867
690 $a 0290
710 2 $a Universitaet Mannheim (Germany). $e degree granting institution. $3 3815696
720 1 $a Gemulla, Rainer $e degree supervisor.
773 0 $t Dissertations Abstracts International $g 86-10A.
790 $a 1124
791 $a Ph.D.
792 $a 2025
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31963743