Supervised classification is one of the most powerful technique to analyze data, when a-priori information is available on the membership of data samples to classes. Unfortunately the labeling process can be both expensive and time-consuming. By exploiting the presence of unlabeled samples, semi-supervised algorithms can produce excellent classification models also when a small amount of labeled data is available. The LapRegec (Laplacian Regec) algorithm is a semi-supervised classifier obtained adding a Laplacian regularization term to Regec. LapRegec produces models that are both accurate and parsimonious in terms of labeled data samples.

M. Viola, M. Sangiovanni, G. Toraldo, and Mario R. Guarracino Semi-supervised generalized eigenvalues classification Annals of Operations Research, 2017.

Presentation

Laplacian-based semi-supervised learning

This work has been funded by MIUR PON02-00619 projects. Mario Guarracino work has been conducted at National Research Institute University Higher School of Economics and has been supported by the RSF grant n. 14-41-00039.