Contact

Professor: Douglas Eck [douglas D0T eck AT umontreal D0T ca] Office Hours: By appointment (3253 Pavillion André-Aisenstadt).

News

07 Jan 2009 The course will be taught at BRAMS on the campus of Université de Montréal. See www.brams.org/contact.html for directions to BRAMS.
01 Jan 2009 Look here ("News") for urgent announcements and read Daily notes for course notes and all other information. The syllabus is available in pdf format.

Teaching

This course will be taught primarily in English and readings will be in English. That said, I'm very happy to have a bilingual environment where each speaks his or her own preferred language. This has worked well in the past. All graded materials (homeworks, final project, presentations) may be done in either English or French

Course Objectives

The course is structured as a graduate seminar with a focus on reading journal papers and implementing machine learning models relevant to music. Our goal is to explore technological challenges in the domain of music from within the framework of machine learning. Areas of investigation include:

There is significant programming work in the homeworks and final project. My goal is that the projects deal with open problems in the field. In the Winter 2005 version of this course two projects were transformed into winning contest submissions in Genre Recognition and Note Onset Detection at an international programming contest (``MIREX'' at the ISMIR 2005 conference).

Evaluation

TaskAmount
Homeworks40%
Weekly Readings10%
Reading Presentations10%
Final Project40%

This is a lab and readings course. There will be no exams.

Homework: (40%) Regular homework assignments dealing with signal processing, music feature extraction and machine learning will be given.

Weekly Readings: (10%) Weekly journal papers, book chapters and tutorial material will be assigned. Careful reading of all material is required.

Reading Presentations: (10%) Each student will present a short summary of one weekly reading and will then lead discussion of the paper.

Final Project: (40%) At mid-term students will submit a 3 to 5 page project proposal for approval. The project must involve an application of machine learning in the domain of music. Students will be furnished with project suggestions and will be shown where to find relevant datasets. Students will present their projects to the class in a 20-30 minute presentation at the end of the term. Example topics:

Prerequisites

Students will be assumed to have either background in machine learning or background in music. For those lacking experience in either domain, extra reading will be required. Note that students having background in both machine learning and music are rare; thus students lacking experience in one of these areas should not be discouraged from taking the course.