Tata Institute of Fundamental Research
Learning to Model: Constraint Acquisition via Queries and Symbolic–Connectionist Hybridization
STCS Seminar
Speaker:
Nadjib Lazaar (University of Paris-Saclay)
Organiser:
Shibashis Guha
Date:
Wednesday, 10 Dec 2025, 16:00 to 17:00
Venue:
HBA Foyer
(Scan to add to calendar)
Abstract:
Constraint Acquisition (CA) is a symbolic learning framework at the intersection of concept learning (inductive reasoning) and constraint programming (deductive reasoning). These two paradigms, rooted in symbolic AI, are combined in CA to automate and accelerate the modeling phase, enabling the automatic construction of constraint networks representing a target concept.
In Query-Based Constraint Acquisition (QBCA), the learner interacts actively with an oracle by posing structured queries, progressively refining the hypothesis space. This talk provides a structured overview of QBCA, with a focus on the typology of queries (e.g., membership queries, equivalence queries) and their role in the acquisition process. I will also cover theoretical foundations, recent algorithmic developments, and extensions to partial, uncertain, or qualitative contexts. QBCA opens promising avenues in applications such as program analysis, autonomous systems, and interactive explainable modeling in XAI.
Finally, I will discuss emerging connections between QBCA and contemporary connectionist approaches, notably language models (LLMs) and Transformers, outlining directions for integrating explicit symbolic reasoning with statistical learning capabilities.
Â
Short Bio:
Nadjib Lazaar is a Full Professor at Paris-Saclay University and a member of the LaHDAK team at the LISN laboratory. He currently serves as director of the MIAGE program and co-director of the Master's in Data Science at the Faculty of Sciences of Orsay. He is also the International Relations Officer for the DATAIA Institute and the AI Cluster. Previously, he was an Associate Professor (HDR) at the University of Montpellier from 2013 to 2024. His research sits at the intersection of Constraint Programming, Data Mining, Machine Learning, and Software Engineering, with a focus on constraint acquisition, declarative data mining, AI-based software testing, and trustworthy neuro-symbolic AI.
| Speaker: | Nadjib Lazaar (University of Paris-Saclay) |
| Organiser: | Shibashis Guha |
| Date: | Wednesday, 10 Dec 2025, 16:00 to 17:00 |
| Venue: | HBA Foyer |
(Scan to add to calendar)
Abstract:
Constraint Acquisition (CA) is a symbolic learning framework at the intersection of concept learning (inductive reasoning) and constraint programming (deductive reasoning). These two paradigms, rooted in symbolic AI, are combined in CA to automate and accelerate the modeling phase, enabling the automatic construction of constraint networks representing a target concept.
In Query-Based Constraint Acquisition (QBCA), the learner interacts actively with an oracle by posing structured queries, progressively refining the hypothesis space. This talk provides a structured overview of QBCA, with a focus on the typology of queries (e.g., membership queries, equivalence queries) and their role in the acquisition process. I will also cover theoretical foundations, recent algorithmic developments, and extensions to partial, uncertain, or qualitative contexts. QBCA opens promising avenues in applications such as program analysis, autonomous systems, and interactive explainable modeling in XAI.
Finally, I will discuss emerging connections between QBCA and contemporary connectionist approaches, notably language models (LLMs) and Transformers, outlining directions for integrating explicit symbolic reasoning with statistical learning capabilities.
Â
Short Bio:
Nadjib Lazaar is a Full Professor at Paris-Saclay University and a member of the LaHDAK team at the LISN laboratory. He currently serves as director of the MIAGE program and co-director of the Master's in Data Science at the Faculty of Sciences of Orsay. He is also the International Relations Officer for the DATAIA Institute and the AI Cluster. Previously, he was an Associate Professor (HDR) at the University of Montpellier from 2013 to 2024. His research sits at the intersection of Constraint Programming, Data Mining, Machine Learning, and Software Engineering, with a focus on constraint acquisition, declarative data mining, AI-based software testing, and trustworthy neuro-symbolic AI.