Description du projet
Améliorer l’IA pour prévoir des événements complexes
De nombreuses applications dépendent de l’IA pour traiter des flux d’informations en constante évolution. Le projet EVENFLOW, financé par l’UE, élaborera des techniques d’apprentissage hybride de prévision des événements complexes en mutation. Le projet aura recours à la fois à des techniques d’apprentissage par représentation neuronale, capables de construire des caractéristiques fondées sur des événements à partir de flux de données au niveau de la perception, et à des outils d’apprentissage et de raisonnement symboliques. La nature en ligne des méthodes d’apprentissage leur permettra d’utiliser les riches connaissances du domaine générées au fil du temps. EVENFLOW sera testé dans le cadre de trois cas: les prévisions oncologiques en médecine de précision, le comportement sûr et efficace des robots de transport autonomes dans les usines intelligentes et l’évaluation fiable du cycle de vie des infrastructures critiques.
Objectif
A growing number of applications rely on AI-based solutions to carry-out mission-critical tasks, many of which are of temporal nature, dealing with ever-evolving flows of information. Crucial for mitigating threats and taking advantage of opportunities in such domains, is the ability to forecast imminent situations and critical complex events ahead of time. EVENFLOW will develop hybrid learning techniques for complex event forecasting, which combine deep learning with logic-based learning and reasoning into neuro-symbolic forecasting models. The envisioned methods will combine (i) neural representation learning techniques, capable of constructing event-based features from streams of perception-level data with (ii) powerful symbolic learning and reasoning tools, that utilize such features to synthesize high-level, interpretable patterns of critical situations to be forecast. Crucial in the EVENFLOW approach is the online nature of the learning methods, which makes them applicable to evolving data flows and allows to utilize rich domain knowledge that is becoming available progressively, over time. To deal with the brittleness of neural predictors and the high volume/velocity of temporal data flows, the EVENFLOW techniques will rely on novel, formal verification techniques for machine learning, in addition to a suite of scalability algorithms for federated training and incremental model construction. The learnt forecasters will be interpretable and scalable, allowing for fully explainable insights, delivered in a timely fashion and enabling proactive decision making. EVENFLOW will be evaluated on three challenging use cases related to oncological forecasting in precision medicine, safe and efficient behavior of autonomous transportation robots in smart factories and reliable life cycle assessment of critical infrastructure.
Champ scientifique
Mots‑clés
Programme(s)
Régime de financement
HORIZON-RIA - HORIZON Research and Innovation ActionsCoordinateur
1050 Bruxelles / Brussel
Belgique