Deep Learning tecniques aplied to semi-supervised problem
Deep Learning, labeling, semi-supervised, clustering, classification.
The expressive growth of modern data sets, combined with the difficulty of obtaining information on labels, has made semi-supervised learning one of the problems of practical importance in modern data analysis. In most cases, obtaining a data set with the amount of examples sufficient to induce a classifier can be costly, because it is necessary that a labeling of the data be performed by an expert. Non-labelled data are easier to obtain, but more difficult to analyze and interpret. In the problem of semi-supervised learning, there is a database consisting of a small labeled part and a larger, unlabeled part. From this, the objective of this work is based on the application of models that use Deep Learning techniques in semi-supervised learning, following two approaches: 1) Semi-supervised classification, where a deep neural network is trained supported by a pre-labeling process, using the labeled data to influence the unlabeled to generate their labels; 2) Semi-supervised grouping, applying a deep grouping by means of an autoencoder network coupled to the K-means algorithm. The labeled data are submitted to the groups generated by the auto encoder, influencing the unlabelled ones, group by group. With the experiments performed, one can notice the satisfactory performance of the proposed models when compared with other semi-supervised algorithms, showing the viability of the models proposed in the problem addressed.