A comparative study of inductive and transductive learning with feedforward neural networks

TitreA comparative study of inductive and transductive learning with feedforward neural networks
Publication TypeJournal Article
Year of Publication2016
AuthorsBianchini, Ma, Belahcen, Ab, Scarselli, Fa
JournalLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10037 LNAI
Pagination283-293
Abstract

Traditional supervised approaches realize an inductive learning process: A model is learnt from labeled examples, in order to predict the labels of unseen examples. On the other hand, transductive learning is less ambitious. It can be thought as a procedure to learn the labels on a training set, while, simultaneously, trying to guess the best labels on the test set. Intuitively, transductive learning has the advantage of being able to directly use training patterns while deciding on a test pattern. Thus, transductive learning faces a simpler problem with respect to inductive learning. In this paper, we propose a preliminary comparative study between a simple transductive model and a pure inductive model, where the learning architectures are based on feedforward neural networks. The goal is to understand how transductive learning affects the complexity (measured by the number of hidden neurons) of the exploited neural networks. Preliminary experimental results are reported on the classical two spirals problem. © Springer International Publishing AG 2016.

URLhttps://www.scopus.com/inward/record.uri?eid=2-s2.0-85006009614&doi=10.1007%2f978-3-319-49130-1_21&partnerID=40&md5=0072365bc5774e0eec549d8f61a2ddc4
DOI10.1007/978-3-319-49130-1_21
Revues: 

Partenaires

Localisation


Location map

Suivez-nous sur

  

Contactez-nous

ENSIAS

Avenue Mohammed Ben Abdallah Regragui, Madinat Al Irfane, BP 713, Agdal Rabat, Maroc

Résultat de recherche d'images pour "icone fax" Télécopie : (+212) 5 37 77 72 30

    Compteur de visiteurs:328,804
    Education - This is a contributing Drupal Theme
    Design by WeebPal.