The MIPS workshop was very interesting for me and for many others. I believe it was a great success. The slides from talks, as well as musical examples, will appear here as soon as I can get them from the authors. For purposes of creating an archive, the information on this page will be left as it was originally written. For example, "NIPS will take place" will not be updated to read "NIPS took place" despite the fact that the workshop has already taken place (21 Dec 2004; DSE).

Welcome to the homepage for MIPS, Music Information Processing Systems Workshop. MIPS will be a one-day workshop to be offered at the Neural Information and Processing Systems (NIPS) 2004 Conference. The main NIPS conference is held from December 13 through 16 in Vancouver. All workshops will be in Whistler, British Columbia, Canada on December 17 and 18 2004 just minutes from the ski slopes. Our workshop will be on December 18th See the NIPS website for information on enrolling in the conference and workshops.

We are no longer accepting proposals for presenting at the MIPS workshop. If you have questions about the workshop please contact Douglas Eck at douglas.eck@umontreal.ca.

Workshop Program

Saturday morning session: 7:30am--10:30am
7:30am:What Can We Learn from Large Music Databases? (Dan Ellis)
8:00am:Phrase Class Models for Music (Darrell Conklin)
8:25am:A Topological Model of Motivic Structure (Chantal Buteau)
8:50am:Bridging Long Timelags in Music (Douglas Eck)
9:15am:Coffee break
9:30am:Panel discussion: Learning Musical Structure
10:00am:Panel discussion: New Directions in Music and Machine Learning

 

Saturday afternoon session: 4:00pm--7:00pm
4:00pm:Score Alignment and Accompaniment Systems (Christopher Raphael)
4:30pm:Generative Models of Music (Ali Taylan Cemgil)
4:55pm:Online Learning for Music Accompaniment (Ofer Dekel, Shai Shalev-Shwartz, Yoram Singer)
5:20pm:Modeling Expressive Musical Performance with HMMs (Graham Grindlay and David Helmbold)
5:45pm:Coffee break
6:00amPanel discussion: Musical Performance and Accompaniment
6:45pm:Closing remarks (D. Eck, D. Ellis, T. Cemgil, J.-F. Paiement)

Workshop Overview

This is the workshop overview published before submissions were accepted. The final program (above) is very close to what we proposed, but there are differences (30 Nov 2004; DSE).

Music raises fundamental issues in machine learning. Musical signals are dynamic and exhibit long term temporal dependencies. Methods for discovering and representing these dependencies will enable the development of algorithms for music analysis and generation. Of particular interest are algorithms that use learned musical structure to compose music and to interact with musicians. Additionally, creating and navigating large databases of music are difficult information retrieval and classification challenge.

The MIPS workshop will explore state-of-the-art machine learning methods in the domain of music. Areas of interest include, but are not limited to:

  • Discovery of musical structure
  • Generative models of music (at signal level and above)
  • Interactive systems
  • Music similarity, classification and retrieval

Format

We propose a one-day workshop with the following structure:
  • brief introduction by workshop organizers
  • five short morning talks
  • morning panel discussion
  • five short afternoon talks
  • afternoon panel discussion
  • brief wrap-up by workshop organizers
In addition to a lunch break, long coffee breaks will be offered both in the morning and afternoon.

Panel discussion topics will be chosen after we have organized our list of presenters. In the tradition of NIPS workshops, our focus is on the discussion of new ideas rather than on creating a conference-like atmosphere. We will actively solicit audience participation throughout the talks and will advise speakers to keep talks short enough so that we have ample time for discussion afterward. In choosing talks we will opt for those that offer new algorithms and challenging ideas --- and also, we hope, interesting musical examples. Submissions will be considered that treat a wide range of music-related machine learning challenges. The workshop favors neither a particular modeling framework nor a particular way of encoding music (from digital audio to symbol strings).

Target Audience

We consider the MIPS workshop to be of interest to most NIPS attendees. Though the workshop is limited to the domain of music, it will touch on a wide range of modeling frameworks and will deal with problems---such as finding long timescale dependencies in a signal and navigating large graphs---that are relevant to those working outside the music domain. Clearly those interested in music should find the workshop interesting. We also believe the workshop will attract researchers from related areas like speech recognition, grammar induction, time series analysis, signal processing and sequence learning.

Call For Papers

The program has been finalized and submissions are no longer being accepted. The Call For Papers is left here for archival reasons (30 Nov 2004; DSE).

The following call-for-papers is also available as an individual text file for printing and circulating. Submissions are due 15 October 2004.

 

  ----------------------------------------------
   MIPS -- Music Information Processing Systems
             Workshop at NIPS*04
      December 17--18, 2004, Whistler, Canada
  ----------------------------------------------

  URL:  http://www.iro.umontreal.ca/~eckdoug/mips

Music raises fundamental issues in machine learning. Musical signals
are dynamic, highly structured and exhibit long term temporal
dependencies. Methods for discovering and representing these
dependencies will enable the development of algorithms for music
analysis and generation.  Of particular interest are algorithms that
use learned musical structure to compose music and to interact with
musicians.  Additionally, creating and navigating large databases of
music are difficult information retrieval and classification
challenge.


Call for Participation
----------------------

We are soliciting participants for the "MIPS -- Music Information
Processing Systems" workshop.  MIPS will concentrate on several key
issues which are currently prominent in the intersection of machine
learning, music information retrieval and computer music research. The
goal is to create an informal platform for scientific discussion of
recent issues in computational methods for music information
processing.  We are especially interested in participants that can
contribute in the following areas:

* Discovery of musical structure
* Generative models of music (at signal and/or symbolic level)
* Interactive systems
* Music similarity, classification and retrieval

Please send an extended abstract not longer than two pages to
Jean-François Paiement at paiement@idiap.ch. PS, PDF and ASCII text are all
acceptable formats. Include the keyword MIPS-SUBMISSION in the title.

The deadline for abstract submission is October 15. Based on these
abstracts, we will select a small number of speakers and a number of panel
members for discussions.

In the tradition of NIPS workshops, the focus is on the discussion of
new ideas rather than on creating a conference-like atmosphere. We
strongly encourage audience participation throughout the talks and will
leave ample time for discussion afterward. In choosing talks we will
opt for those that offer novel algorithms, challenging ideas and
interesting musical examples. Submissions will be
considered that treat a wide range of music-related machine learning
challenges.  The workshop favors neither a particular modeling
framework nor a particular way of encoding music (from digital audio
to symbol strings).

The Workshop
-------------

MIPS is a one-day workshop with a morning and afternoon session.
The tentative schedule is:

* brief introduction by workshop organizers
* five short morning talks
* morning panel discussion
* five short afternoon talks
* afternoon panel discussion
* brief wrap-up by workshop organizers

In addition to a lunch break, long coffee breaks will be offered both in the
morning and afternoon.

Additionally, there will be a possibility to present posters and
demonstrations during these breaks.

NIPS (http://nips.cc) is the leading conference on machine learning,
applied statistics, cognitive modeling and
neurocomputation. Following the main conference in Vancouver, the
workshops are held at Whistler Ski Resort and combine wonderful
science and skiing in an informal environment.

Workshop Organizers
-------------------

Douglas Eck
University of Montreal, Department of Computer Science
Montreal, Quebec, Canada
douglas.eck@umontreal.ca

Daniel P. W. Ellis
Dept. of Electrical Engineering, Columbia University
New York, NY
dpwe@ee.columbia.edu

Ali Taylan Cemgil
IAS Informatica Instituut, University of Amsterdam
The Netherlands
cemgil@science.uva.nl

Jean-François Paiement
IDIAP Research Institute
Martigny, Switzerland
paiement@idiap.ch

Last modified: Tue Dec 21 08:33:27 EST 2004 by Douglas Eck [douglas D0T eck AT umontreal D0T ca]