Société Française de Bioinformatique

Mots-Clés

AI artificial intelligence logic programming XAI symbolic AI

Description

Project Overview
This research internship is part of the LOGIX project at INRAE and is intended for students interested in pursuing research in artificial intelligence. Depending on the intern’s performance, research interests, and project needs, the internship may lead to a PhD position within the same research project. The LOGIX project develops methods for extracting interpretable, logic-based explanations from neural networks and applying them to biomedical prediction tasks. The research combines explainable AI (XAI), formal verification, answer set programming (ASP), and machine learning to improve the transparency and reliability of AI systems in high-stakes domains such as healthcare.
Research Activities
The successful candidate will contribute to several aspects of the project, including:
• Studying explainable AI and formal verification methods for neural networks.
• Developing logic-based encodings for neural network explanations.
• Extracting interpretable rules from biomedical prediction models.
• Implementing and evaluating explanation methods on benchmark and biomedical datasets.
• Assessing the robustness, reliability, and scalability of proposed approaches.
• Contributing to open-source software and, where appropriate, scientific publications.
The internship combines theoretical research with practical implementation and experimental evaluation.

Candidate Profile
Applicants should be pursuing a degree in:
• Computer Science
• Artificial Intelligence
• Mathematics
• Data Science
• Applied Logic or a related field
The ideal candidate should have:
• A strong interest in AI, machine learning, and explainability.
• Programming experience (preferably Python).
• Good analytical and problem-solving skills.
• Good written and spoken English.
Experience with logic programming, symbolic AI, or formal verification is beneficial but not required. Motivation to conduct research and learn new concepts is more important.

Selected References
1. ASP-based explanation approach for AI models
https://link.springer.com/chapter/10.1007/978-3-032-25305-7_1
2. Thrombosis prediction using machine learning
https://www.nature.com/articles/s41598-021-93390-7
3. Risk of recurrence in thrombosis
https://link.springer.com/chapter/10.1007/978-3-030-85633-5_7