| Acronym: NEXUS |
| Main Objective: This research project aims to advance causality and explainability in machine learning by integrating principles from information theory, abductive reasoning, and geometric approaches from physics. Addressing challenges such as time-series complexities, various biases, algorithmic sensitivity, and the need for transparent, reproducible explanations, the project adopts an interdisciplinary strategy to improve interpretability, generalization, and computational efficiency. By leveraging concepts like minimum description length, Kolmogorov complexity, and geometric invariants, it seeks to create simpler, more robust models and novel methods for understanding feature importance and causal dependencies. The release of open-source tools highlights its commitment to transparency and community engagement. Ultimately, this ambitious initiative aims to set new standards for innovation in causality and explainability within machine learning. |
| Reference: UID/50008/2025 - Instituto de Telecomunicações |
| Funding: FCT |
| Approval Date: 20-11-2025 |
| Start Date: 21-11-2025 |
| End Date: 20-11-2027 |
| Team: Alexandra Sofia Martins de Carvalho, Mario Alexandre Teles de Figueiredo, Paulo Alexandre Carreira Mateus, Maria Cristina de Sales Viana Serôdio Sernadas, João Filipe Quintas dos Santos Rasga, Bruno Miguel Santos Mera |
| Groups: Pattern and Image Analysis – Lx, Security and Quantum Information - Lx |
| Partners: IT |
| Local Coordinator: Alexandra Sofia Martins de Carvalho |
|
|
