Deep Learning for Sound and Music Processing Workshop

Table of contents:

The workshop presentations were recorded on video. If you are interested in receiving a copy of the recordings, please send a request to Maarten Grachten (e-mail address on: http://www.ofai.at/~maarten.grachten/).


Workshop summary

Topic: Deep Learning for Sound and Music Processing
Location: OFAI, Vienna
Data: December 4-5, 2014 (Thu-Fri)
Fee: Participation is free, capacity is limited
Registration deadline: November 25, 2014
URL: http://lrn2cre8.eu/?q=deeplearningworkshop
Contact: maarten.grachten@ofai.at


Call

Dear colleagues,

You are kindly invited to participate in the workshop, "Deep Learning for Sound and Music Processing", to be held at the Austrian Research Institute for Artificial Intelligence (OFAI), in Vienna, on the 4th and 5th of December 2014 (Thu and Fri).

The primary aim of the workshop is to share knowledge and experience in the deployment of deep learning architectures and algorithms, particularly in the context of music related research. In addition to presentations of music research involving deep learning, attention will be paid to "lrn2", a python framework for representation learning from data, which was developed in the context of the European Project "Learning to create". More information about lrn2 can be found online: http://lrn2cre8.eu/?q=workpackage1/deliverable1.1

The workshop consists of a seminar part (thursday) and a tutorial part (friday). The seminar will include a brief introduction to deep learning methods, several presentations of sound and music processing research involving deep learning, and an overview of the lrn2 software framework for representation learning. The tutorial is optional, and is targeted at participants that wish to get started with the application of deep learning methods. It takes the form of a hands-on session, where participants can use the software. This may be as simple as running and inspecting some demonstration source code, but also allows participants to employ deep learning methods for their own purposes, by modifing existing or writing new code. At the end of the session, there will be an opportunity for participants to present results, and discuss their experiences.

For participation in the tutorial, some basic programming experience is required. Some knowledge about the python programming language would be helpful. Mathematical skills are useful for understanding the models implemented in the framework, but are not required for basic usage. Furthermore, participants are expected to bring their own laptop for the tutorial. Code execution can take place on OFAI computers through ssh logins, and program editing can be done by participants locally on their laptops, via network shares. The only requirements on the participants' laptops are therefore an ssh-client, network share mounting functionality, and a text editor.

Participation in the workshop is free of charge, but registration by email is required becaused of limited seating. Address your request for participation to: maarten.grachten@ofai.at (please specify whether you wish to attend just the seminar, or both the seminar and the tutorial). Coffee/tea will be served during the workshop, but lunch is not included. Topical contributions to the seminar are welcome: please contact us at your earliest opportunity in case you should like to apply for a presentation slot.


Agenda

A tentative agenda for the workshop is below:

Thursday December 4th

10:00 Introduction to Deep Learning (Maarten Grachten)

Recurrent Neural Networks:

10:30 Beat Tracking with LSTM-RNNs (Sebastian Böck)
11:00 Hessian-Free Learning (Carlos Cancino)

11:30 Coffee Break

(Stacked) Autoencoders

12:00 Learning binary codes for fast music retrieval (Jan Schlüter)

(Stacked) Restricted Boltzmann Machines

12:30 Speech/Music classification (Jan Schlüter)

13:00 Lunch break

14:15 Learning tonal structure from melodies (Carlos Cancino)

Convolutional Neural Networks and Dropout

14:45 Onset detection / Audio segmentation (Jan Schlüter)

15:15 High-dimensional aspects of CNNs (Karen Ullrich)

15:45 Coffee Break

Lrn2Cre8 software framework

16:00 Introduction to lrn2 framework (Maarten Grachten)

19:00 Punschparty at OFAI 2

Friday December 5th

9:00 - 13:00
* Installation / Setup of lrn2 software framework for participants
* Usage demonstration of lrn2
* Participants present their ideas/needs
* Participants work on their projects

Lunch

14:00 - 17:00
* Participants work on their projects
* Participants present any results and discuss their experiences

We are looking forward to your participation.

Kindest regards,

Maarten Grachten (OFAI),
Stefan Lattner (OFAI),
Kat Agres (QMUL),
Carlos Eduardo Cancino Chacón (OFAI)

Austrian Research Institute for Artificial Intelligence (OFAI)
Freyung 6/6, A-1010 Vienna, Austria
Phone: (+43-1) 5336112-0
Fax: (+43-1) 5336112-77


How to get there

An elaborate description of how to get to OFAI is available on the OFAI website. The workshop will take place in OFAI1.


Accommodation

Hotel Regina
Tel. +43 1 404 46-0
http://www.kremslehnerhotels.at/de/hotel-regina-wien/
Special rates for OFAI guests:
Single room: € 90,- including breakfast

Pension Liechtenstein
http://www.pensionliechtenstein.at/

Arcotel Boltzmann
http://www.hotelboltzmann.at/
This hotel is for real Deep Learning fanatics :-)

Bleckmann Hotel
http://www.hotelbleckmann.at/

Austria Trend Hotel Rathauspark
reservierung.rathauspark@austria-trend.at
http://www.austria-trend.at/

Hotel Atlanta
http://www.hirners.com/

Pension Franz Hotel
http://www.vho.at/franz/index.de.html

Pension Excellence
http://www.pension-excellence.com/

Hotel Pension Astra
http://www.hotelpensionastra.com/
Single rooms for 40-50 euros