Tuning Active Sampling Techniques for Evolutionary Learner from Big Data Sets: Review and Discussion

Sana Ben Hamida , Marta Rukoz (1)

1 LAMSADE - Laboratoire d'analyse et modélisation de systèmes pour l'aide à la décision

Big data processing is the new challenge for analytical, machine learning techniques. Many efforts are needed to scale both classic, advanced methods to the the mass of the provided data. Evolutionary learning algorithms (EAL) are robust, effective methods in solving a wide variety of complex learning problems. This paper discusses how to tune the active sampling techniques for EAL to deal with very large training data sets. It introduces the key decisions needed to design an effective active sampling strategy, review the main techniques used with evolutionary algorithms. Then, we discuss how they could be adapted to learn from big training data sets, present some research directions in this domain.

Mots clés: Training; Training data; Evolutionary computation; Supervised learning; Data mining; Sociology; Statistics

Domaines: Informatique [cs]

Sana Ben Hamida, Marta Rukoz. Tuning Active Sampling Techniques for Evolutionary Learner from Big Data Sets: Review and Discussion. UIC/ATC/ScalCom/CBDCom/IoP/SmartWorld (2016 Intl IEEE Conferences), Jul 2016, Toulouse, France. pp.1206-1213, ⟨10.1109/UIC-ATC-ScalCom-CBDCom-IoP-SmartWorld.2016.0184⟩. ⟨hal-01448255⟩