AN ADAPTIVE SYSTEM FOR MUSIC CLASSIFICATION AND TAGGING (MIREX 2009 SUBMISSION)

摘要: This extended abstract concerns one of the two systems submitted by IRCAM for participation in MIREX 2009 classification and tagging tasks. The system is adaptive can handle both single-label tasks (genre, mood, artist) multilabel (tagging). Adaptability attained means automatic feature model selection, which are embedded multiple-instance binary relevance learning a Support Vector Machine. We propose criterion function SVM parameter selection that takes into account unbalanced sets effects overfitting. same algorithm, without any manual adaptation, was to all However, it evaluated different configurations (also tasks) related temporal modeling methods: first mode (“file”) each track represented single vector second (“tw”) texture windows fixed length computed, with later decision fusion.