ESI 2017
Systematic approaches to deep learning methods for audio
Monday, September 11. - Friday, September 15. 2017


Monday, September 11.


 09:00 - 09:30 :  
Registration and Opening

 09:30 - 10:40 :  Jan Schlueter:  Deep learning as an engineer: The nuts and bolts and dirty tricks

 10:45 - 11:15 :  
Coffee Break

 11:15 - 12:25 :  Philipp Grohs:  Deep learning as a mathematician: Conjectures, proofs and open questions

 12:30 - 14:30 :  
Lunch

 14:30 - 15:15 :  Fabio Anselmi:  Invariant and selective representations with applications to Deep learning
 15:20 - 16:15 :  Emmanuel Vincent:  When mismatched training data outperform matched data

 16:15 - 16:40 :  
Wrap-up


 17:00 - 19:00 :  
Welcome-Gathering



Tuesday, September 12.

 09:30 - 10:20 :  Irene Waldspurger:  Inversion of the wavelet transform modulus

 10:30 - 11:00 :  
Coffee Break

 11:00 - 11:35 :  Joakim Andén:  Joint Time-Frequency Scattering Networks
 11:45 - 12:20 :  Vincent Lostanlen:  The spiral and the snowball: deeper enhancements to time-frequency scattering

 12:45 - 14:30 :  
Lunch

 14:45 - 15:15 :  Roswitha Bammer:  Invariance and Stability of Gabor Scattering for Music Signals
 15:30 - 16:05 :  Antoine Deleforge:  Reversed Mixture-of-Experts Networks for High- to Low-Dimensional Mapping

 16:15 - 16:45 :  
Wrap-up



Wednesday, September 13.

 09:30 - 10:15 :  Simon Durand:  Deep learning for downbeat tracking of music audio signals
 10:15 - 10:40 :  Stefan Balke:  Literature Review: Deep Neural Networks in MIR

 10:45 - 11:15 :  
Coffee Break

 11:25 - 11:55 :  Pavol Harar:  Voice Pathology Detection Using Deep Learning: a Preliminary Study
 12:00 - 12:35 :  Aggelos Gkiokas:  Causal Time Series Processing with Convolutional Neural Networks. An application to Real Time Beat Tracking

 12:45 - 14:30 :  
Lunch

 14:30 - 15:15 :  Karen Ullrich:  Bayesian Neural Networks: Techniques and Applications
 15:15 - 15:50 :  Andre Holzapfel:  Bayesian meter tracking on learned signal representations

 16:15 - 16:45 :  
Wrap-up



Thursday, September 14.

 09:30 - 10:15 :  Guido Montufar:  Exponential advantages of deep and distributed representations

 10:15 - 10:45 :  
Coffee Break

 10:45 - 11:30 :  Grégoire Montavon:  Explaining the Predictions of Deep Neural Networks
 12:00 - 12:35 :  Hendrik Vincent Koops:  Learning Shared Chord Representations for Annotator Subjectivity

 12:45 - 14:25 :  
Lunch

 14:30 - 15:05 :  Sebastian Stober:  Automatic Speech Recognition (on a Budget) - Transfer Learning and Introspection
 15:10 - 15:40 :  Olga Slizovskaia:  Correspondence between audio and visual deep models for musical instrument detection in video recordings
 15:40 - 16:15 :  Mishra Saumitra:  Interpretable machine learning for music content analysis

 16:15 - 16:45 :  
Wrap-up


 19:00 - 22:00 :  
Workshop dinner



Friday, September 15.

 09:30 - 10:10 :  Matthias Dorfer:  Learning Correspondences between Audio and Sheet-Music Images
 10:10 - 10:50 :  Oriol Nieto:  Cold-Start Music Recommendation Using Multimodal Deep Architectures

 10:30 - 11:00 :  
Coffee Break

 11:00 - 11:55 :  Sander Dielemann:  Deep learning for music recommendation and generation

 12:00 - 13:00 :  
Wrap-up and Closing