Probabilistic Graphical Models for the Analysis of Musical Audio
Details
This month we're continuing with so much awesome for the NYCML group. Matt Hoffman, a postdoc at Columbia, will be presenting "Probabilistic Graphical Models for the Analysis of Musical Audio". Here's Matt's abstract and bio.
Content-based Music Information Retrieval (MIR) systems seek to automatically extract meaningful information from musical audio signals. I will discuss applications of new and existing generative probabilistic models to several content-based MIR tasks: timbral similarity estimation, semantic annotation and retrieval, and latent source discovery and separation.
In order to estimate how similar two songs sound to one another, we employ a Hierarchical Dirichlet Process (HDP) mixture model to discover a shared representation of the distribution of timbres in each song. Comparing songs under this shared representation yields better query-by-example retrieval quality and scalability than previous approaches.
To predict what tags are likely to apply to a song (e.g., “rap,” “happy,” or “driving music”), we develop the Codeword Bernoulli Average (CBA) model, a simple and fast mixture-of-experts model. Despite its simplicity, CBA performs at least as well as previous state-of-the-art approaches at automatically annotating songs and finding to what songs in a database a given tag most applies.
Finally, we address the problem of latent source discovery and separation by developing a Bayesian nonparametric model, gamma process nonnegative matrix factorization (GaP-NMF). GaP-NMF allows us to discover what sounds (e.g. bass drums, guitar chords, etc.) are present in a song and to isolate or suppress individual sources. Crucially, this model is able to decide how many latent sources are necessary to model the data. This feature is particularly valuable in this application, since it is impossible to guess a priori how many sounds will appear in a given song.
Bio: Matt Hoffman is a postdoc working in the department of Statistics at Columbia University. He holds a B.S. in Computer Science from Columbia University, a B.A. in Management-Engineering from Claremont McKenna College, and an M.A. and Ph.D. in Computer Science from Princeton University.
His research interests are in the areas of machine learning, probabilistic modeling and approximate inference, Bayesian nonparametric modeling, computer and human audition, music information retrieval, and audio synthesis.