An introduction to data mining through the lens of music information retrieval. Topics explored include classification (genre, mood, instrument), multi-label classification (tagging), and regression (emotion/mood).
Session 1: Naive Bayes Classification In this session, we will learn about the main idea of generative classifiers using probabilistic modeling, Bayes theorem, the naive bayes assumption, evaluation of classification, cross-validation.
Session 2: Discriminating Classifiers Decision trees, perceptron, artificial neural networks, support vector machines will be covered in this session.
Session 3: Tagging This session is about methods of tag acquisition (surveys, games with a purpose), auto-tagging architectures, evaluation of auto-tagging.
Session 4: Regression We will learn about Regression and how it is applied in emotion/mood recognition, and other regression applications such as surrogate sensing for music instruments.