Support Vector Machine Active Learning for Music Mood Tagging

Active learning is a subfield of machine learning based on the idea that the accuracy of an algorithm can be improved with fewer training samples if it is allowed to choose the data from which it learns. We present the results for Support Vector Machine (SVM) active learning experiments for music mood tagging based on a multi-sample selection strategy that chooses samples according to their proximity to the boundary, their proximity to points in the training set and the density around them. The influence of those key active learning parameters is assessed by means of ANalysis Of Variance (ANOVA). Using these analyses we demonstrate the efficiency of active learning compared to typical fulldataset batch learning: our method allows to tag music by mood more efficiently than a regular approach, requiring fewer instances to obtain the same performance than using random sample selection methods.