Knowledge-driven active learning - Laboratoire Jean-Alexandre Dieudonné Accéder directement au contenu
Pré-Publication, Document De Travail Année : 2022

Knowledge-driven active learning

Résumé

In the last few years, Deep Learning models have become increasingly popular. However, their deployment is still precluded in those contexts where the amount of supervised data is limited and manual labelling expensive. Active learning strategies aim at solving this problem by requiring supervision only on few unlabelled samples, which improve the most model performances after adding them to the training set. Most strategies are based on uncertain sample selection, and even often restricted to samples lying close to the decision boundary. Here we propose a very different approach, taking into consideration domain knowledge. Indeed, in the case of multi-label classification, the relationships among classes offer a way to spot incoherent predictions, i.e., predictions where the model may most likely need supervision. We have developed a framework where first-order-logic knowledge is converted into constraints and their violation is checked as a natural guide for sample selection. We empirically demonstrate that knowledge-driven strategy outperforms standard strategies, particularly on those datasets where domain knowledge is complete. Furthermore, we show how the proposed approach enables discovering data distributions lying far from training data. Finally, the proposed knowledge-driven strategy can be also easily used in object-detection problems where standard uncertainty-based techniques are difficult to apply.
Fichier principal
Vignette du fichier
2110.08265.pdf (5.03 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03526510 , version 1 (14-01-2022)

Identifiants

  • HAL Id : hal-03526510 , version 1

Citer

Gabriele Ciravegna, Frédéric Precioso, Marco Gori. Knowledge-driven active learning. 2022. ⟨hal-03526510⟩
36 Consultations
17 Téléchargements

Partager

Gmail Facebook X LinkedIn More