Dynamische Probabilistische Relationale Modelle
Projektleiter: Prof. Dr. Ralf Möller
Wissenschaftlicher Mitarbeiter: Marcel Gehrke
In praktischen Anwendung ist es für Agenten wichtig, über eine große Menge von Objekten sinnvoll schlussfolgern zu können (und nicht nur über ein Objekt oder wenige). Dabei sollte erreicht werden, dass nicht jedes Objekt einzeln betrachtet werden muss, sondern für Gruppen von Objekten mit gleichen Eigenschaften jeweils Stellvertreter betrachtet werden können. In den Anwendung ist Information über einzelne Objekte mit Unsicherheit belegt, so dass Schlussfolgerungen über Unsicherheiten unterstützt werden müssen, und auch das zugrundeliegende Modell berücksichtigt Unsicherheiten.
Information über einzelne Objekte werden vom Agenten über der Zeit akquiriert, so dass auch zeitliches Schließen eine große Bedeutung in pratischen Anwendungen hat. Wenn nun aber Evidenz über einzelne Objekte über der Zeit eintrifft, so wird es immer weniger möglich, für Gruppen von Objekten effizient über Stellvertreter zu schließen, so dass Schlussfolgerungsprozesse zur Berechnung von optimalen Handlungen immer langsamer werden. In der Praxis sollte also über der Zeit wieder von "Einzeleindrücken" abstrahiert werden, ohne zu große Fehler für Einzelobjekte zu erzeugen.
Die Veröffentlichung löst das zeitliche Informationsabstraktionsproblem zum ersten Male im Rahmen von temporalen (dynamischen) probabilistischen relationalen Modellen (DPRMs) und stellt damit einen wichtigen Schritt zur stabilen Datenverarbeitung in einem Agenten dar, so dass Handlungen auch mit fortschreitend eintreffender Information über verschiedene Objekte effizient berechnet werden können und vom Agenten trotzdem viele Einzelobjekte "im Blick behalten werden können"
Im Rahmen von COPICOH haben wir DPRMs im Bereich Medizin und Gesundheit angwendet. Im DFG-Exzellencluster UWA modellieren wir semantische Repräsentationen von Texten mit DPRMs.
Lifted Dynamic Junction Tree Algorithm
We work on probabilistic first-order formalisms where the domain objects are known. In these formalisms, the standard approach for inference with first-order constructs include lifted variable elimination (LVE) for single queries. To handle multiple queries efficiently and to obtain a compact representation, the lifted junction tree algorithm (LJT) extends LVE. We extend the formalism and respectively LJT to handle temporal aspects. To be more precise, we combine the advantages of LJT and the interface algorithm in LDJT, which efficiently solves the inference problems filtering and prediction.
Additionally, we are interested in solving other inference problems, e.g. smoothing, and to learn relational temporal models from data.
LDJT is supported by CISCO. The work is carried out as part of Jointlab 1 within the COPICOH center for connected health.
LDJT für Textverstehen
Im DFG-Exzellenzcluster "Understanding Written Artefacts" setzen wir LDJT ein, um semantisches Repräsentationslernen zu studieren.
Implementation
A prototype implementation of LDJT based on BLOG and the LVE implementation by Taghipour as well as some documentation is available:
The web pages around the implementation have been prepared by Moritz Hoffmann.
Publications
2023
- Magnus Bender, Tanya Braun, Ralf Möller, Marcel Gehrke: LESS is More: LEan Computing for Selective Summaries
wird veröffentlicht in: KI 2023: Advances in Artificial Intelligence, 2023, Springer Nature Switzerland, p.1-14
- Magnus Bender, Kira Schwandt, Ralf Möller, Marcel Gehrke: FrESH – Feedback-reliant Enhancement of Subjective Content Descriptions by Humans
wird veröffentlicht in: Proceedings of the Workshop on Humanities-Centred Artificial Intelligence (CHAI 2023), , CEUR Workshop Proceedings - Malte Luttermann, Ralf Möller, Marcel Gehrke: Lifting Factor Graphs with Some Unknown Factors
in: Proceedings of the Seventeenth European Conference on Symbolic and Quantitative Approaches to Reasoning with Uncertainty (ECSQARU-23), 2023, Springer - Florian Andreas Marwitz, Ralf Möller, Marcel Gehrke: PETS: Predicting Efficiently using Temporal Symmetries in Temporal PGMs
in: Proceedings of the Seventeenth European Conference on Symbolic and Quantitative Approaches to Reasoning with Uncertainty (ECSQARU-23), 2023, Springer - Magnus Bender, Tanya Braun, Ralf Möller, Marcel Gehrke: Unsupervised Estimation of Subjective Content Descriptions in an Information System
in: International Journal of Semantic Computing, 2023 - Magnus Bender, Tanya Braun, Ralf Möller, Marcel Gehrke: Unsupervised Estimation of Subjective Content Descriptions
in: 17th IEEE International Conference on Semantic Computing, (ICSC 2023), February 1-3, 2023, IEEE
2022
- Marcel Gehrke, Ralf Möller, Tanya Braun: Who did it? Identifying the Most Likely Origins of Events
in: Proceedings of the 11th International Conference on Probabilistic Graphical Models (PGM 2022), 2022, p.217-228 - Tanya Braun, Marcel Gehrke: Explainable and Explorable Decision Support
in: Proceedings of the 27th International Conference on Conceptual Structures (ICCS 2022), Münster, Germany, September 12-15, 2022, 2022 - Tanya Braun, Marcel Gehrke, Florian Lau, Ralf Möller: Lifting in Multi-agent Systems under Uncertainty
in: 38th Conference on Uncertainty in Artificial Intelligence (UAI 2022), Eindhoven, Netherlands, August 1-5, 2022, 2022
2021
- Magnus Bender, Tanya Braun, Marcel Gehrke, Felix Kuhr, Ralf Möller, Simon Schiff: Identifying and Translating Subjective Content Descriptions Among Texts
in: International Journal of Semantic Computing, 2021, Vol.15, (4), p.461-485 - Marcel Gehrke: Taming Exact Inference in Temporal Probabilistic Relational Models
University of Lübeck, 2021, PhD thesis - Marcel Gehrke: On the Complexity and Completeness of the Lifted Dynamic Junction Tree Algorithm
in: 10th International Workshop on Statistical Relational AI at the 1st International Joint Conference on Learning and Reasoning, 2021 - Tanya Braun, Marcel Gehrke, Tom Hanika, Nathalie Hernandez (Eds.): ICCS-21 Proceedings of the 26th International Conference on Conceptual Structures
Springer, 2021 - Nils Finke, Tanya Braun, Marcel Gehrke, Ralf Möller: Concept Drift Detection in Dynamic Probabilistic Relational Models
in: The International FLAIRS Conference Proceedings, 2021, Vol.34 - Nils Finke, Tanya Braun, Marcel Gehrke, Ralf Möller: Dynamic Domain Sizes in Temporal Probabilistic Relational Models
in: The International FLAIRS Conference Proceedings, 2021, Vol.34 - Magnus Bender, Tanya Braun, Marcel Gehrke, Felix Kuhr, Ralf Möller, Simon Schiff: Identifying Subjective Content Descriptions Among Texts
in: 15th IEEE International Conference on Semantic Computing, (ICSC 2021), Laguna Hills, CA, USA, January 27-29, 2021, IEEE, p.9-16
2020
- Nils Finke, Marcel Gehrke, Tanya Braun, Tristan Potten, Ralf Möller: Investigating Matureness of Probabilistic Graphical Models for Dry-Bulk Shipping
in: Proceedings of the 10th International Conference on Probabilistic Graphical Models, 2020, 23-25 Sep, Manfred Jaeger, Thomas Dyhre Nielsen volu (Ed.), PMLR, Proceedings of Machine Learning Research, p.197-208 - Marcel Gehrke, Tanya Braun, Simon Polovina: Restricting the Maximum Number of Actions for Decision Support under Uncertainty
in: ICCS-20 Proceedings of the 25th International Conference on Conceptual Structures, 2020 - Marcel Gehrke, Ralf Möller, Tanya Braun: Taming Reasoning in Temporal Probabilistic Relational Models
in: Proceedings of the 24th European Conference on Artificial Intelligence (ECAI 2020), 2020, p.2592 - 2599 - Stefan Lüdtke, Marcel Gehrke, Tanya Braun, Ralf Möller, Thomas Kirste: Lifted Marginal Filtering for Asymmetric Models by Clustering-based Merging
in: Proceedings of the 24th European Conference on Artificial Intelligence (ECAI 2020), 2020 - Marcel Gehrke, Tanya Braun, Ralf Möller: Taming Reasoning in Temporal Probabilistic Relational Models
in: 9th International Workshop on Statistical Relational AI at the 34th AAAI Conference on Artificial Intelligence, 2020
2019
- Marcel Gehrke, Tanya Braun, Ralf Möller: Efficient Multiple Query Answering in Switched Probabilistic Relational Models
in: Proceedings of AI 2019: Advances in Artificial Intelligence, 2019, Springer, Lecture Notes in Computer Science, Vol.11919, p.104-116 - Marcel Gehrke, Simon Schiff, Tanya Braun, Ralf Möller: Which Patient to Treat Next? Probabilistic Stream-based Reasoning for Decision Support and Monitoring
in: Proceedings of the ICBK 2019, 2019, IEEE, p.73-80 - Mattis Hartwig, Marcel Gehrke, Ralf Möller: Approximate Query Answering in Complex Gaussian Mixture Models
in: Proceedings of the ICBK 2019, 2019, IEEE, p.81-86 - Marcel Gehrke, Tanya Braun, Ralf Möller: Lifted Temporal Most Probable Explanation
in: Graph-Based Representation and Reasoning - 24th International Conference on Conceptual Structures, (ICCS 2019), Marburg, Germany, July 1-4,, 2019, Springer, Lecture Notes in Computer Science, Vol.11530, p.72-85 - Tanya Braun, Marcel Gehrke: Inference in Statistical Relational AI
in: Proceedings of the International Conference on Conceptual Structures 2019, 2019, Springer, p.xvii-xix - Marcel Gehrke, Tanya Braun, Ralf Möller: Lifted Temporal Maximum Expected Utility
in: Proceedings of the 32nd Canadian Conference on Artificial Intelligence, Canadian AI 2019, 2019, Springer, p.380-386 - Marcel Gehrke, Tanya Braun, Ralf Möller: Uncertain Evidence for Probabilistic Relational Models
in: Proceedings of the 32nd Canadian Conference on Artificial Intelligence, Canadian AI 2019, 2019, Springer, Lecture Notes in Computer Science, Vol.11489, p.80-93 - Marcel Gehrke, Tanya Braun, Ralf Möller: Relational Forward Backward Algorithm for Multiple Queries
in: Proceedings of the 32nd International Florida Artificial Intelligence Research Society Conference (FLAIRS-19), 2019, AAAI Press, p.464-469 - Marcel Gehrke, Tanya Braun, Ralf Möller, Alexander Waschkau, Christoph Strumann, Jost Steinhäuser: Lifted Maximum Expected Utility
in: Artificial Intelligence in Health, 2019, Springer International Publishing, p.131-141
2018
- Marcel Gehrke, Tanya Braun, Ralf Möller: Answering Multiple Conjunctive Queries with the Lifted Dynamic Junction Tree Algorithm
in: Proceedings of the AI 2018: Advances in Artificial Intelligence, 2018, Springer, p.543-555 - Marcel Gehrke, Tanya Braun, Ralf Möller: Preventing Unnecessary Groundings in the Lifted Dynamic Junction Tree Algorithm
in: Proceedings of the AI 2018: Advances in Artificial Intelligence, 2018, Springer, p.556-562 - Simon Schiff, Marcel Gehrke, Ralf Möller: Efficient Enriching of Synthesized Relational Patient Data with Time Series Data
in: Procedia Computer Science, 2018, Vol.141, p.531 - 538, The 8th International Conference on Current and Future Trends of Information and Communication Technologies in Healthcare (ICTH-2018) / Affiliated Workshops - Marcel Gehrke, Tanya Braun, Ralf Möller: Towards Preventing Unnecessary Groundings in the Lifted Dynamic Junction Tree Algorithm
in: Proceedings of KI 2018: Advances in Artificial Intelligence, 2018, Springer, p.38-45 - Marcel Gehrke, Tanya Braun, Ralf Möller: Preventing Unnecessary Groundings in the Lifted Dynamic Junction Tree Algorithm
in: 8th International Workshop on Statistical Relational AI at the 27th International Joint Conference on Artificial Intelligence, 2018 - Marcel Gehrke, Tanya Braun, Ralf Möller: Answering Hindsight Queries with Lifted Dynamic Junction Trees
in: 8th International Workshop on Statistical Relational AI at the 27th International Joint Conference on Artificial Intelligence, 2018 - Marcel Gehrke, Tanya Braun, Ralf Möller, Alexander Waschkau, Christoph Strumann, Jost Steinhäuser: Towards Lifted Maximum Expected Utility
in: Proceedings of the First Joint Workshop on Artificial Intelligence in Health in Conjunction with the 27th IJCAI, the 23rd ECAI, the 17th AAMAS, and the 35th ICML, 2018, CEUR-WS.org, CEUR Workshop Proceedings, Vol.2142, p.93-96 - Marcel Gehrke, Tanya Braun, Ralf Möller: Lifted Dynamic Junction Tree Algorithm
in: Proceedings of the International Conference on Conceptual Structures, 2018, Springer, p.55-69