===================================================== Learning Models for Interactive Melodic Improvisation ===================================================== - Belinda Thom My research addresses the problem of a computer interacting with a live, improvising musician in the jazz/blues setting. First, I introduce BoB (Band-out-of-a-Box), a model whose function is trading solos with a musician in a customized, adaptive manner. Next, I introduce a novel EM-based unsupervised learning algorithm that allows BoB to unobtrusively train itself, by merely watching the musician improvise. I demonstrate the power of this approach by presenting the abstractions that emerge when Charlie Parker's Mohawk solos are analyzed. As a bonus, the main body of the talk will be followed by a discussion of some of the details of the novel EM algorithm utilized.