On Data-Preparation Efficiency Application on Breast Cancer Classification

Mouna Sabrine Mayouf (1) , Florence Dupin de Saint-Cyr (1, 2)

1 IRIT-ADRIA - Argumentation, Décision, Raisonnement, Incertitude et Apprentissage
2 UT3 - Université Toulouse III - Paul Sabatier

Quantifying the informative state of a dataset for a classification task is an important question. In this paper, we try to circumscribe this notion by introducing some new measures and enunciating some principles about data preparation. We experiment the interest of these measures and the validity of these principles by introducing several protocols designed for comparing different ways to prepare the data. We conclude by relating the efficiency of the data preparation and its theoretical diversity.

Mots clés: deep learning; convolutional neural networks; data preparation; information metrics; Breakhis

Domaines: Intelligence artificielle [cs.AI]; Imagerie médicale

Mouna Sabrine Mayouf, Florence Dupin de Saint-Cyr. On Data-Preparation Efficiency Application on Breast Cancer Classification. IRIT/RR–2020–09–FR, IRIT : Institut de Recherche en Informatique de Toulouse. 2020. ⟨hal-03246076⟩