I am now a Research Scientist at Google Research in Mountain View. As of June 1, 2011 I have resigned my tenured position as Associate Professor at University of Montreal. Though I continue to be active in the department as an Adjunct Professor, I am no longer accepting new students.
The best contact email for me is douglas.<last_name>@gmail.com.
I will update this page with a redirect to a new page soon. The research links on this page are still valid (papers, etc) but some details are wrong.
Cool...Here are some cool 3D views of my Boesendorfer lab and also the anechoic chamber at BRAMS. The photos were taken by photographer Thomas Chevet:
ResearchHere is a new and improved summary of the Search Inside the Music music recommendation project.
For non-research types: A few months ago I described my research to an old friend from high school. My wife later read the email and said it was pretty easy to understand. So here it is.
When I was at Univeristy of Montreal, I spent half of my time at LISA, the
University of Montreal Montreal machine learning lab. I spent the other half of my time at the Brain, Music and Sound (BRAMS)
research center affiliated with the University of Montreal, McGill
University and the Montreal Neurological Institute (MNI).
Here's a picture of the Boesendorfer piano in my lab at BRAMS with lab technician (and electroacoustic composer) Dominic Thibault. Click on the image for a higer-resolution version:
My overall area of research is machine learning in the domain of music. Specific research areas (past and present) include:
- Measuring music similarity and recommending music (see the SITM description page)
- Music sequence generation. (Here are some examples using LSTM recurrent neural networks)
- Expression: Realistic Performance Timing and Dynamics
- Tracking: Finding Beat and Tempo in a Musical Performance
- Synchronization: Playing with Musicians in Real Time
See also my tech talk at Google on Automatically Tagging Audio Files Using Supervised Learning on Acoustic Features (Video)
- IFT1025 (Autumn 2008) Programming II (Programmation II)
- IFT6080 (Winter 2009) Music and Machine Learning (Musique et Apprentissage Automatique)
- IFT3205 (Winter 2009) Signal Processing (Traitement de Signal)
News and MiscellanyICML 2009 Workshop: Sparse Methods for Music Audio With Dan Ellis (Associate Professor, Electrical Engineering, Columbia University) and Philippe Hamel (PhD Candidate in our group), I organized a workshop focusing on Sparse Methods for Music Audio.
Montreal Music and Machine Learning Workshop 2008 The the first Montreal Music and Machine Learning Workshop at BRAMS was held on November 14th. See the web page for details. It was a great success. Thanks to the sponsors:
I spent 6 months working with Sun Labs researcher Paul Lamere (see below). I realized today that I had never posted a link to Paul Lamere's EchoNest blog. So here it is: Music Machinery.
Search Inside the Music (SITM): Navigating Tag Space
Here is a new and improved summary of the Search Inside the Music music recommendation project.
Good news for BRAMS
In January 2007 the BRAMS group received a 14M$ grant from Canada Fund for Innovation.
I have some thoughts concerning randomness and repetition in music.
Apropos musical similarity, I have some versions of My Favorite Things to listen to.
Our (U.S.) President's Particular Appetites
You can't make this stuff up.
In Canada, Colgate sells children's toothpaste that plays music. You open the cap and it plays a 70 second song ("Yankee Doodle Dandy"). The idea is that your child will use that time to brush effectively. What is strange is that they transpose the second part of the song (beginning "Yankee Doodle keep it up"), generating an entirely new version. The only reason I can think of for this transposition is that it makes the range of notes smaller. In fact, the new version fits into one octave on the keyboard while the original needs more than one octave. Perhaps this limitation was necessary to get a particular "toothpaste cap note generating device" to be sufficiently cheap. Or maybe they just thought it sounded better. (Who can say? The toothpaste tube features the same Barney the Dinosaur whose shows are used to torture Iraqi prisoners). The MIDI file YankeeDoodleToothpaste.mid contains the original Yankee Doodle melody played once, followed by a rendition of the Toothpaste version. Note that this is just my MIDI recreation of the two versions: there is no way to download a MIDI file from a toothpaste cap. Though I matched the tempo from memory, the MIDI version is not a full 70 seconds long. This is because the toothpaste tube plays the first part of the song three times before going into the (transposed) second part whereas I only play it once to avoid putting you to sleep. PS I've grown to prefer the Toothpaste version after multiple (forced) listenings.
Music and Machine Learning Workshop at NIPS*04
Dan Ellis, Taylan Cemgil, Jean-François Paiement and myself organized a workshop devoted to music and machine learning at NIPS 2004, Saturday December 18 in Whistler, Canada as part of the two-day workshop series Dec 17 and 18 2004. The workshop was called Music Information Processing Systems (MIPS). It was very interesting for me and for many others. I will be contacting authors soon to send their slides in PDF or Powerpoint format and I will put them online. See the MIPS homepage for details.
Rhythm Perception and Production Workshop (RPPW)
I was co-organizer with Carolyn Drake and Sophie Scott of the 2003 Rhythm Perception and Production Workshop (RPPW) in Ile de Tatihou, Normandy, France. I also edited a special issue of Music Perception dedicated to that meeting.
VIBE Oscillator Simulator
For something a bit older, have a look at my java-based oscillator simulator. That can be found at the Vibe homepage.
Octave Mailing List searcher
My Octave mailing list searcher has now been incorporated into the Octave site.
Academic BiographyUniversity of Montreal (Present): I am an associate professor in the University of Montreal Department of Computer Science and Operations Research. I'm affiliated with the Adaptive Systems Laboratory (LISA), whose members include Yoshua Bengio and Balazs Kegl and many others. I'm a founding member of the Montreal Center for Brain, Music and Sound, BRAMS, a research center devoted to the study of music cognition (among other things). This group is directed by Isabelle Peretz of the University of Montreal and Robert Zatorre of the Montreal Neurological Institute (MNI). (P.S. Isabelle thinks "founding member" is a funny phrase. Too strong? Too close to "foundering member"?)
IDSIA (2000 to 2003): I worked as a postdoctoral researcher at The Dalle Molle Institute for Artificial Intelligence (IDSIA) in Lugano, Switzerland. My post-doctoral director was Juergen Schmidhuber. While at IDSIA I focused on music composition and music improvisation using hybrid recurrent neural networks, as well as recognition of phonemes in continuous speech. A large portion of my research concerned the Long Short-Term Memory (LSTM) model developed by Sepp Hochreiter while a student of Juergen Schmidhuber. I worked with Felix Gers, Bram Bakker, Alex Graves, Ivo Kwee and Marcus Hutter. Special mention to Marcus for being the only person I know with a music-playing pocket calculator web site.
Indiana University (Dissertation 2000): I did my PhD at Indiana University in Computer Science and Cognitive Science. My research focused on musical rhythm perception and production using ensembles of biologically-realistic spiking neurons. My thesis director was Mike Gasser and I worked with (other) nice people like Bob Port, Linda B. Smith, Brian Cantwell Smith, Tony Chemero, Devin McAuley and Fred Cummins.
PersonalWhen I'm not working...
- I play with my wife Amie, my daughter Olivia and my son Samuel.
- I play jazz piano, bluegrass guitar. In fact, you can listen to some of my beautiful (?) music.
- I play ultimate
University of Montreal
Department of Computer Science and Operations Research
CP 6128, Succ. Centre-Ville
H3C 3J7 CANADA
University of Montreal
Department of Computer Science and Operations Research
Pavillon Andre-Aisenstadt, Room 3253 (dept. office 2194)
2920 chemin de la Tour
H3T 1J4 CANADA
Email firstname [dot] lastname [at] umontreal.ca (hint: use my full first name, "Douglas")|
Voice +1 (514) 343-6111 ext 3520.
Note: our "state-of-the-art" phone system won't actually connect you with 3520. Instead you must enter 1-3520-#. Go figure.
Fax +1 (514) 343-5834
A text-file version of contact information (to copy and paste) is available here.