%0 Conference Proceedings
%4 sid.inpe.br/sibgrapi/2018/08.31.09.43
%2 sid.inpe.br/sibgrapi/2018/08.31.09.43.59
%@doi 10.1109/SIBGRAPI.2018.00057
%T Semi-Supervised Learning with Interactive Label Propagation guided by Feature Space Projections
%D 2018
%A Benato, Bárbara Caroline,
%A Telea, Alexandru Cristian,
%A Falcão, Alexandre Xavier,
%@affiliation University of Campinas
%@affiliation University of Groningen
%@affiliation University of Campinas
%E Ross, Arun,
%E Gastal, Eduardo S. L.,
%E Jorge, Joaquim A.,
%E Queiroz, Ricardo L. de,
%E Minetto, Rodrigo,
%E Sarkar, Sudeep,
%E Papa, João Paulo,
%E Oliveira, Manuel M.,
%E Arbeláez, Pablo,
%E Mery, Domingo,
%E Oliveira, Maria Cristina Ferreira de,
%E Spina, Thiago Vallin,
%E Mendes, Caroline Mazetto,
%E Costa, Henrique Sérgio Gutierrez,
%E Mejail, Marta Estela,
%E Geus, Klaus de,
%E Scheer, Sergio,
%B Conference on Graphics, Patterns and Images, 31 (SIBGRAPI)
%C Foz do Iguaçu, PR, Brazil
%8 29 Oct.-1 Nov. 2018
%I IEEE Computer Society
%J Los Alamitos
%S Proceedings
%K Semi-Supervised Learning, Interactive Label Propagation, Auto-Encoder Neural Networks, Visual Analytics.
%X While the number of unsupervised samples for data annotation is usually high, the absence of large supervised train- ing sets for effective feature learning and design of high-quality classifiers is a known problem whenever specialists are required for data supervision. By exploring the feature space of supervised and unsupervised samples, semi-supervised learning approaches can usually improve the classification system. However, these approaches do not usually exploit the pattern-finding power of the users visual system during machine learning. In this paper, we incorporate the user in the semi-supervised learning process by letting the feature space projection of unsupervised and supervised samples guide the label propagation actions of the user to the unsupervised samples. We show that this procedure can significantly reduce user effort while improving the quality of the classifier on unseen test sets. Due to the limited number of supervised samples, we also propose the use of auto-encoder neural networks for feature learning. For validation, we compare the classifiers that result from the proposed approach with the ones trained from the supervised samples only and semi-supervised trained using automatic label propagation.
%@language en
%3 PID5546009.pdf