Graph data management
Expressive query languages, recursive navigational queries, optimization, and scalable graph processing.
Research
Modern AI increasingly relies on data that is structured, interconnected, and governed by rich semantics. A central challenge is to design methods that exploit this relational structure for querying, learning, and reasoning, while combining statistical learning with symbolic reasoning methods.
In the Tyrex project-team, we conduct research at the interface of data management, artificial intelligence, and formal methods. Our work spans expressive query languages, recursive graph querying, relational learning, and neuro-symbolic approaches that integrate logic with machine learning. We study these questions from their algebraic and logical foundations to scalable open-source systems and applications on real-world data.
Research themes
Relational and neuro-symbolic learning
Learning from structured data by combining relational models, statistical methods, and symbolic knowledge.
Languages and reasoning
Logical and algebraic foundations for querying, verification, inference, and hybrid reasoning systems.
Projects
GraphRec
Recursive graph queries · Relational learning · Hybrid querying
I coordinate the ANR-funded GraphRec project, in which we investigate recursive navigational graph query processing, and hybrid approaches to querying large graphs. Recursive navigational graph queries enable the exploration of paths of arbitrary length within graph-structured data, allowing efficient retrieval of relationships that are not explicitly stored but can be inferred through recursion. Additionally, we study relational learning methods to extract meaningful patterns from relational data.
CLEAR
Large-scale data processing · Code generation · Predictive models
Through the ANR-funded CLEAR project, we studied language and compilation techniques for large-scale data processing. Our work addressed graph querying, the generation of optimized distributed programs for platforms such as Apache Spark, and scalable methods for learning predictive models from massive datasets.