Analytically tractable sample-specific confidence measures for semi-supervised learning

In general, classifiers require a large number of labelled training samples preferably covering a wide range of variation to yield a comprehensive and generalising recognition behaviour. However, manual labelling of huge data sets is costly and time-consuming, which is the main motivation for the development of semi-supervised learning algorithms which autonomously extend the “knowledge” gained by classifiers based on a small amount of initial, manually labelled training samples towards increasingly different representatives of the regarded pattern classes.