@INPROCEEDINGS{leonard+sikora+defrancisco+eck:2010,
AUTHOR = {B. Leonard and G. Sikora and M. De Francisco and D. Eck},
TITLE = {Acoustic Space Sampling and the Grand Piano in a
Non-Anechoic Environment: a recordist-centric approach to
to musical acoustic study},
BOOKTITLE = {129th Audio Engineering Society (AES) Convention},
YEAR = {2010},
ADDRESS = {London},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference}
}
@INPROCEEDINGS{courville+eck+bengio:nips2009,
AUTHOR = {A. Courville and D. Eck and Y. Bengio},
TITLE = {An Infinite Factor Model Hierarchy Via a Noisy-Or
Mechanism},
YEAR = {2010},
BOOKTITLE = {Neural Information Processing Systems Conference 22
(NIPS'09)},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
EDITOR = {},
PUBLISHER = {},
PDF = {},
NOTE = {Accepted.}
}
@INPROCEEDINGS{davies+plumbley+eck:waspaa2009,
AUTHOR = {M. Davies and M. Plumbley and D. Eck},
TITLE = {Towards a musical beat emphasis function},
BOOKTITLE = {Proceedings of IEEE WASPAA},
YEAR = {2009},
ADDRESS = {New Paltz, NY},
ORGANIZATION = {IEEE Workshop on Applications of Signal Processing to
Audio and Acoustics},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference}
}
@INPROCEEDINGS{hamel+wood+eck:ismir2009,
AUTHOR = {P. Hamel and S. Wood and D. Eck},
TITLE = {Automatic identification of instrument classes in
polyphonic and poly-instrument audio},
YEAR = {2009},
BOOKTITLE = {{Proceedings of the 10th International Conference on Music
Information Retrieval ({ISMIR} 2009)}},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
EDITOR = {},
PUBLISHER = {}
}
@INPROCEEDINGS{maillet+eck+desjardins+lamere:ismir2009,
AUTHOR = {F. Maillet and D. Eck and G. Desjardins and P. Lamere},
TITLE = {Steerable Playlist Generation by Learning Song Similarity
from Radio Station Playlists},
YEAR = {2009},
BOOKTITLE = {{Proceedings of the 10th International Conference on Music
Information Retrieval ({ISMIR} 2009)}},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
EDITOR = {},
PUBLISHER = {}
}
@ARTICLE{paiement+bengio+eck:aij,
AUTHOR = {{J.-F.} Paiement and S. Bengio and D. Eck},
TITLE = {Probabilistic Models for Melodic Prediction},
JOURNAL = {Artificial Intelligence Journal},
YEAR = {2009},
SOURCE = {OwnPublication},
SOURCETYPE = {Journal},
VOLUME = {173},
PAGES = {1266-1274}
}
@ARTICLE{bertinmahieux+eck+maillet+lamere:jnmr2008,
AUTHOR = {T. Bertin-Mahieux and D. Eck and F. Maillet and P.
Lamere},
TITLE = {Autotagger: A Model For Predicting Social Tags from
Acoustic Features on Large Music Databases},
JOURNAL = {Journal of New Music Research},
YEAR = {2008},
VOLUME = {37},
NUMBER = {2},
PAGES = {115--135},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2008_jnmr.pdf},
SOURCE = {OwnPublication},
SOURCETYPE = {Journal}
}
@TECHREPORT{eck+lapalme:2008,
AUTHOR = {D. Eck and J. Lapalme},
TITLE = {Learning Musical Structure Directly from Sequences of
Music},
INSTITUTION = {Universit\'e de Montr\'eal DIRO},
YEAR = {2008},
NUMBER = {1300},
ADDRESS = {http://www.iro.umontreal.ca/\-\~{}eckdoug/papers/tr1300.pdf},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/tr1300.pdf},
SOURCE = {OwnPublication},
SOURCETYPE = {TechReport}
}
@INPROCEEDINGS{manzagol+bertinmahieux+eck:ismir2008,
AUTHOR = {P-A. Manzagol and T. Bertin-Mahieux and D. Eck},
TITLE = {On the use of Sparse Time Relative Auditory Codes for
Music},
YEAR = {2008},
BOOKTITLE = {{Proceedings of the 9th International Conference on Music
Information Retrieval ({ISMIR} 2008)}},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
EDITOR = {},
PUBLISHER = {},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2008_ismir.pdf}
}
@UNPUBLISHED{kegl+bertinmahieux+eck:mml2008,
AUTHOR = {B. K{\'e}gl and T. Bertin-Mahieux and D. Eck},
TITLE = {{Metropolis-Hastings} Sampling in a {FilterBoost} Music
Classifier},
YEAR = {2008},
NOTE = {International Workshop on Machine Learning and Music
(ICML/COLT/UAI 2008), Helsinki, Finland},
SOURCE = {OwnPublication},
SOURCETYPE = {Workshop},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2008_mml.pdf}
}
@INPROCEEDINGS{paiement+grandvalet+bengio+eck:icml2008,
AUTHOR = {{J.-F.} Paiement and Y. Grandvalet and S. Bengio and D.
Eck},
TITLE = {A generative model for rhythms},
BOOKTITLE = {ICML '08: Proceedings of the 25th International Conference
on Machine Learning},
YEAR = {2008},
PAGES = {},
LOCATION = {Helsinki, Finland},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference}
}
@UNPUBLISHED{eck:nipsworkshop2007,
AUTHOR = {D. Eck},
TITLE = {Measuring and modeling musical expression},
NOTE = {NIPS 2007 Workshop on Music, Brain and Cognition},
SOURCE = {OwnPublication},
SOURCETYPE = {Workshop},
OPTKEY = {},
OPTMONTH = {},
YEAR = {2007},
OPTANNOTE = {}
}
@UNPUBLISHED{paiement+grandvalet+bengio+eck:nipsworkshop2007,
AUTHOR = {{J.-F.} Paiement and Y. Grandvalet and S. Bengio and D.
Eck},
TITLE = {A generative model for rhythms},
NOTE = {NIPS 2007 Workshop on Music, Brain and Cognition},
SOURCE = {OwnPublication},
SOURCETYPE = {Workshop},
OPTKEY = {},
OPTMONTH = {},
YEAR = {2007},
OPTANNOTE = {}
}
@UNPUBLISHED{pugin+burgoyne+eck+fujinaga:nipsworkshop2007,
AUTHOR = {L. Pugin and J.A. Burgoyne and D. Eck and I. Fujinaga},
TITLE = {Book-adaptive and book-dependant models to accelerate
digitalization of early music},
NOTE = {NIPS 2007 Workshop on Music, Brain and Cognition},
SOURCE = {OwnPublication},
SOURCETYPE = {Workshop},
OPTKEY = {},
OPTMONTH = {},
YEAR = {2007},
OPTANNOTE = {}
}
@INPROCEEDINGS{eck+lamere+bertinmahieux+green:nips2007,
AUTHOR = {D. Eck and P. Lamere and T. Bertin-Mahieux and S. Green},
TITLE = {Automatic generation of social tags for music
recommendation},
YEAR = {2008},
BOOKTITLE = {Neural Information Processing Systems Conference 20
(NIPS'07)},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
EDITOR = {},
PUBLISHER = {},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2007_nips.pdf}
}
@INPROCEEDINGS{eck+bertinmahieux+lamere:ismir2007,
AUTHOR = {D. Eck and T. Bertin-Mahieux and P. Lamere},
TITLE = {Autotagging music using supervised machine learning},
YEAR = {2007},
BOOKTITLE = {{Proceedings of the 8th International Conference on Music
Information Retrieval ({ISMIR} 2007)}},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
EDITOR = {},
PUBLISHER = {},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2007_ismir.pdf}
}
@INPROCEEDINGS{lamere+eck:ismir2007,
AUTHOR = {P. Lamere and D. Eck},
TITLE = {Using 3D Visualizations to Explore and Discover Music},
YEAR = {2007},
BOOKTITLE = {{Proceedings of the 8th International Conference on Music
Information Retrieval ({ISMIR} 2007)}},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
EDITOR = {},
PUBLISHER = {}
}
@INCOLLECTION{jaeger+eck:2007,
AUTHOR = {H. Jaeger and D. Eck},
TITLE = {Can't get you out of my head: {A} connectionist model of
cyclic rehearsal},
BOOKTITLE = {{Modeling Communications with Robots and Virtual Humans}},
SERIES = {{LNCS}},
PUBLISHER = {Springer-Verlag},
YEAR = {2007},
SOURCE = {OwnPublication},
SOURCETYPE = {Chapter},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2007_jaeger_eck.pdf}
}
@INPROCEEDINGS{eck:icassp2007,
AUTHOR = {D. Eck},
TITLE = {Beat Tracking Using an Autocorrelation Phase Matrix},
YEAR = {2007},
BOOKTITLE = {{Proceedings of the 2007 International Conference on
Acoustics, Speech and Signal Processing (ICASSP)}},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
EDITOR = {},
PAGES = {1313--1316},
PUBLISHER = {IEEE Signal Processing Society},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2007_icassp.pdf}
}
@INPROCEEDINGS{bergstra+lacoste+eck:ismir2006,
AUTHOR = {J. Bergstra and A. Lacoste and D. Eck},
TITLE = {Predicting genre labels for artists using FreeDB},
BOOKTITLE = {{Proceedings of the 7th International Conference on Music
Information Retrieval ({ISMIR} 2006)}},
YEAR = {2006},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
PAGES = {85-88},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2006_ismir_freedb.pdf}
}
@ARTICLE{eck:mp2006,
AUTHOR = {D. Eck},
TITLE = {Finding Long-Timescale Musical Structure with an
Autocorrelation Phase Matrix},
YEAR = {2006},
JOURNAL = {Music Perception},
VOLUME = {24},
NUMBER = {2},
PAGES = {167--176},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2006_rppw.pdf},
SOURCE = {OwnPublication},
SOURCETYPE = {Journal}
}
@ARTICLE{lacoste+eck:eurasip,
AUTHOR = {A. Lacoste and D. Eck},
TITLE = {A Supervised Classification Algorithm For Note Onset
Detection},
JOURNAL = {EURASIP Journal on Applied Signal Processing},
YEAR = {2007},
VOLUME = {2007},
NUMBER = {ID 43745},
PAGES = {1--13},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2006_eurasip_draft.pdf},
SOURCE = {OwnPublication},
SOURCETYPE = {Journal}
}
@ARTICLE{bergstra+casagrande+erhan+eck+kegl:ml,
AUTHOR = {J. Bergstra and N. Casagrande and D. Erhan and D. Eck and
B. K{\'e}gl},
TITLE = {Aggregate Features and {AdaBoost} for Music
Classification},
JOURNAL = {Machine Learning},
YEAR = {2006},
VOLUME = {65},
NUMBER = {2-3},
PAGES = {473-484},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2006_ml_draft.pdf},
SOURCE = {OwnPublication},
SOURCETYPE = {Journal}
}
@INPROCEEDINGS{eck:icmpc2006,
AUTHOR = {D. Eck},
TITLE = {Beat Induction Using an Autocorrelation Phase Matrix},
YEAR = {2006},
BOOKTITLE = {The Proceedings of the 9th International Conference on
Music Perception and Cognition ({ICMPC9})},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
PAGES = {931-932},
EDITOR = {M. Baroni and A. R. Addessi and R. Caterina and M. Costa},
PUBLISHER = {Causal Productions}
}
@INPROCEEDINGS{paiement+eck+bengio:ccai2006,
AUTHOR = {{J.-F.} Paiement and D. Eck and S. Bengio},
TITLE = {Probabilistic Melodic Harmonization},
BOOKTITLE = {Canadian Conference on AI},
YEAR = {2006},
PAGES = {218-229},
EDITOR = {Luc Lamontagne and Mario Marchand},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
PUBLISHER = {Springer},
SERIES = {Lecture Notes in Computer Science},
VOLUME = {4013}
}
@MISC{eck+scott:editor2005,
AUTHOR = {D. Eck and S. K. Scott},
TITLE = {Music Perception},
YEAR = 2005,
NOTE = {Guest Editor, Special Issue on Rhythm Perception and
Production, 22(3)},
SOURCE = {OwnPublication},
SOURCETYPE = {Other}
}
@ARTICLE{eck+scott:2005,
AUTHOR = {D. Eck and S. K. Scott},
JOURNAL = {Music Perception},
TITLE = {Editorial: New Research in Rhythm Perception and
Production},
YEAR = 2005,
VOLUME = {22},
NUMBER = {3},
PAGES = {371-388},
SOURCE = {OwnPublication},
SOURCETYPE = {Other}
}
@INPROCEEDINGS{paiement+eck+bengio+barber:icml2005,
AUTHOR = {{J.-F.} Paiement and D. Eck and S. Bengio and D. Barber},
TITLE = {A graphical model for chord progressions embedded in a
psychoacoustic space},
BOOKTITLE = {ICML '05: Proceedings of the 22nd international conference
on Machine learning},
YEAR = {2005},
PAGES = {641--648},
LOCATION = {Bonn, Germany},
PUBLISHER = {ACM Press},
ADDRESS = {New York, NY, USA},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference}
}
@INPROCEEDINGS{casagrande+eck+kegl:icmc2005,
AUTHOR = {N. Casagrande and D. Eck and B. Kegl},
TITLE = {Geometry in Sound: A Speech/Music Audio Classifier
Inspired by an Image Classifier},
BOOKTITLE = {{Proceedings of the International Computer Music
Conference (ICMC)}},
YEAR = {2005},
PAGES = {207--210},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2005_icmc_casagrande.pdf},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference}
}
@UNPUBLISHED{eck:rppw2005,
AUTHOR = {D. Eck},
TITLE = {Meter and Autocorrelation},
NOTE = {{10th Rhythm Perception and Production Workshop (RPPW),
Alden Biesen, Belgium}},
YEAR = {2005},
SOURCE = {OwnPublication},
SOURCETYPE = {Workshop}
}
@INPROCEEDINGS{paiement+eck+bengio:ismir2005,
AUTHOR = {{J.-F.} Paiement and D. Eck and S. Bengio},
TITLE = {A Probabilistic Model for Chord Progressions},
BOOKTITLE = {{Proceedings of the 6th International Conference on Music
Information Retrieval ({ISMIR} 2005)}},
YEAR = {2005},
PAGES = {312-319},
ADDRESS = {London: University of London},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference}
}
@INPROCEEDINGS{eck+casagrande:ismir2005,
AUTHOR = {D. Eck and N. Casagrande},
TITLE = {Finding Meter in Music Using an Autocorrelation Phase
Matrix and Shannon Entropy},
BOOKTITLE = {{Proceedings of the 6th International Conference on Music
Information Retrieval ({ISMIR} 2005)}},
YEAR = {2005},
PAGES = {504--509},
ADDRESS = {London: University of London},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2005_ismir.pdf}
}
@UNPUBLISHED{eck:nipsworkshop2006,
AUTHOR = {D. Eck},
TITLE = {Generating music sequences with an echo state network},
NOTE = {NIPS 2006 Workshop on Echo State Networks and Liquid State
Machines},
SOURCE = {OwnPublication},
SOURCETYPE = {Workshop},
OPTKEY = {},
OPTMONTH = {},
YEAR = {2006},
ABSTRACT = {Slides and musical examples available on request.},
OPTANNOTE = {}
}
@INPROCEEDINGS{casagrande+eck+kegl:ismir2005,
AUTHOR = {N. Casagrande and D. Eck and B. K\'{e}gl},
TITLE = {Frame-Level Audio Feature Extraction using {A}da{B}oost},
BOOKTITLE = {{Proceedings of the 6th International Conference on Music
Information Retrieval ({ISMIR} 2005)}},
YEAR = {2005},
PAGES = {345--350},
ADDRESS = {London: University of London},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2005_ismir_casagrande.pdf}
}
@UNPUBLISHED{mirex2005genre,
AUTHOR = {J. Bergstra and N. Casagrande and D. Eck},
TITLE = {Genre Classification: Timbre- and Rhythm-Based
Multiresolution Audio Classification},
NOTE = {{MIREX} genre classification contest},
SOURCE = {OwnPublication},
SOURCETYPE = {Other},
ADDRESS = {{ISMIR} Conference, London},
YEAR = {2005}
}
@UNPUBLISHED{mirex2005artist,
AUTHOR = {J. Bergstra and N. Casagrande and D. Eck},
TITLE = {Artist Recognition: A Timbre- and Rhythm-Based
Multiresolution Approach},
NOTE = {{MIREX} artist recognition contest},
SOURCE = {OwnPublication},
SOURCETYPE = {Other},
ADDRESS = {{ISMIR} Conference, London},
YEAR = {2005}
}
@UNPUBLISHED{mirex2005note,
AUTHOR = {A. Lacoste and D. Eck},
TITLE = {Onset Detection with Artificial Neural Networks},
NOTE = {{MIREX} note onset detection contest},
SOURCE = {OwnPublication},
SOURCETYPE = {Other},
ADDRESS = {{ISMIR} Conference, London},
YEAR = {2005}
}
@UNPUBLISHED{mirex2005tempo,
AUTHOR = {D. Eck and N. Casagrande},
TITLE = {A Tempo-Extraction Algorithm Using an Autocorrelation
Phase Matrix and Shannon Entropy},
NOTE = {{MIREX} tempo extraction contest
(www.music-ir.org/\-evaluation/\-mirex-results)},
SOURCE = {OwnPublication},
SOURCETYPE = {Other},
ADDRESS = {{ISMIR} Conference, London},
YEAR = {2005}
}
@UNPUBLISHED{eck:mipsworkshop2004,
AUTHOR = {D. Eck},
TITLE = {Bridging Long Timelags in Music},
NOTE = {NIPS 2004 Workshop on Music and Machine Learning (MIPS),
Whistler, British Columbia},
SOURCE = {OwnPublication},
SOURCETYPE = {Workshop},
OPTKEY = {},
OPTMONTH = {},
YEAR = {2004},
ABSTRACT = {Slides and musical examples available on request.},
OPTANNOTE = {}
}
@UNPUBLISHED{eck:bramsworkshop2004,
AUTHOR = {D. Eck},
TITLE = {Challenges for Machine Learning in the Domain of Music},
NOTE = {BRAMS Workshop on Brain and Music, Montreal Neurological
Institute},
SOURCE = {OwnPublication},
SOURCETYPE = {Workshop},
OPTKEY = {},
OPTMONTH = {},
YEAR = {2004},
ABSTRACT = {Slides and musical examples available on request.},
OPTANNOTE = {}
}
@INPROCEEDINGS{eck:icmpc2004,
AUTHOR = {D. Eck},
TITLE = {A Machine-Learning Approach to Musical Sequence Induction
That Uses Autocorrelation to Bridge Long Timelags},
YEAR = {2004},
ADDRESS = {Adelaide},
BOOKTITLE = {{The Proceedings of the Eighth International Conference on
Music Perception and Cognition ({ICMPC}8)}},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
EDITOR = {S.D. Lipscomb and R. Ashley and R.O. Gjerdingen and P.
Webster},
PUBLISHER = {Causal Productions},
PAGES = {542-543},
ABSTRACT = { One major challenge in using statistical sequence
learning methods in the domain of music lies in bridging
the long timelags that separate important musical events.
Consider, for example, the chord changes that convey the
basic structure of a pop song. A sequence learner that
cannot predict chord changes will almost certainly not be
able to generate new examples in a musical style or to
categorize songs by style. Yet, it is surprisingly
difficult for a sequence learner to bridge the long
timelags necessary to identify when a chord change will
occur and what its new value will be. This is the case
because chord changes can be separated by dozens or
hundreds of intervening notes. One could solve this problem
by treating chords as being special (as did Mozer, NIPS
1991). But this is impractical---it requires chords to be
labeled specially in the dataset, limiting the
applicability of the model to non-labeled examples---and
furthermore does not address the general issue of nested
temporal structure in music. I will briefly describe this
temporal structure (known commonly as "meter") and present
a model that uses to its advantage an assumption that
sequences are metrical. The model consists of an
autocorrelation-based filtration that estimates online the
most likely metrical tree (i.e. the frequency and phase of
beat, measure, phrase &etc.) and uses that to generate a
series of sequences varying at different rates. These
sequences correspond to each level in the hierarchy.
Multiple learners can be used to treat each series
separately and their predictions can be combined to perform
composition and categorization. I will present preliminary
results that demonstrate the usefulness of this approach.
Time permitting I will also compare the model to alternate
approaches. }
}
@INPROCEEDINGS{graves+eck+schmidhuber:bio-adit2004,
AUTHOR = {A. Graves and D. Eck and N. Beringer and J. Schmidhuber},
TITLE = {Biologically Plausible Speech Recognition with {LSTM}
Neural Nets},
YEAR = {2004},
BOOKTITLE = {Proceedings of the First Int'l Workshop on Biologically
Inspired Approaches to Advanced Information Technology
(Bio-ADIT)},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2004_bioadit.pdf},
PAGES = {127-136},
ABSTRACT = { Long Short-Term Memory (LSTM) recurrent neural networks
(RNNs) are local in space and time and closely related to a
biological model of memory in the prefrontal cortex. Not
only are they more biologically plausible than previous
artificial RNNs, they also outperformed them on many
artificially generated sequential processing tasks. This
encouraged us to apply LSTM to more realistic problems,
such as the recognition of spoken digits. Without any
modification of the underlying algorithm, we achieved
results comparable to state-of-the-art Hidden Markov Model
(HMM) based recognisers on both the TIDIGITS and TI46
speech corpora. We conclude that LSTM should be further
investigated as a biologically plausible basis for a
bottom-up, neural net-based approach to speech recognition.
}
}
@ARTICLE{perez+gers+schmidhuber+eck:2003,
AUTHOR = {J.A. P\'{e}rez-Ortiz and F. A. Gers and D. Eck and J.
Schmidhuber},
TITLE = {{K}alman filters improve {LSTM} network performance in
problems unsolvable by traditional recurrent nets},
JOURNAL = {Neural Networks},
NUMBER = 2,
VOLUME = 16,
PAGES = {241--250},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2003_nn.pdf},
YEAR = {2003},
SOURCE = {OwnPublication},
SOURCETYPE = {Journal},
ABSTRACT = {The Long Short-Term Memory (LSTM) network trained by
gradient descent solves difficult problems which
traditional recurrent neural networks in general cannot. We
have recently observed that the decoupled extended Kalman
filter training algorithm allows for even better
performance, reducing significantly the number of training
steps when compared to the original gradient descent
training algorithm. In this paper we present a set of
experiments which are unsolvable by classical recurrent
networks but which are solved elegantly and robustly and
quickly by LSTM combined with Kalman filters.}
}
@UNPUBLISHED{eck:nipsworkshop2003,
AUTHOR = {D. Eck},
TITLE = {Time-warped hierarchical structure in music and speech: A
sequence prediction challenge},
NOTE = {NIPS 2003 Workshop on Recurrent Neural Networks, Whistler,
British Columbia},
SOURCE = {OwnPublication},
SOURCETYPE = {Workshop},
OPTKEY = {},
OPTMONTH = {},
YEAR = {2003},
ABSTRACT = {Slides and musical examples available on request.},
OPTANNOTE = {}
}
@UNPUBLISHED{eck:irisworkshop2004,
AUTHOR = {D. Eck},
TITLE = {Using Autocorrelation to Bridge Long Timelags when
Learning Sequences of Music},
NOTE = {IRIS 2004 Machine Learning Workshop, Ottawa, Canada},
SOURCE = {OwnPublication},
SOURCETYPE = {Workshop},
OPTKEY = {},
OPTMONTH = {},
YEAR = {2004},
ABSTRACT = {Slides and musical examples available on request.},
OPTANNOTE = {}
}
@TECHREPORT{eck+graves+schmidhuber:tr-speech2003,
AUTHOR = {D. Eck and A. Graves and J. Schmidhuber},
TITLE = {A New Approach to Continuous Speech Recognition Using
{LSTM} Recurrent Neural Networks},
INSTITUTION = {IDSIA},
YEAR = {2003},
NUMBER = {IDSIA-14-03},
ADDRESS = {www.idsia.ch/\-techrep.html},
MONTH = {May},
SOURCE = {OwnPublication},
SOURCETYPE = {TechReport},
ABSTRACT = { This paper presents an algorithm for continuous speech
recognition built from two Long Short-Term Memory (LSTM)
recurrent neural networks. A first LSTM network performs
frame-level phone probability estimation. A second network
maps these phone predictions onto words. In contrast to
HMMs, this allows greater exploitation of long-timescale
correlations. Simulation results are presented for a
hand-segmented subset of the "Numbers-95" database. These
results include isolated phone prediction, continuous
frame-level phone prediction and continuous word
prediction. We conclude that despite its early stage of
development, our new model is already competitive with
existing approaches on certain aspects of speech
recognition and promising on others, warranting further
research. }
}
@TECHREPORT{graves+eck+schmidhuber:tr-digits2003,
AUTHOR = {A. Graves and D. Eck and J. Schmidhuber},
TITLE = {Comparing {LSTM} Recurrent Networks and Spiking Recurrent
Networks on the Recognition of Spoken Digits},
INSTITUTION = {IDSIA},
YEAR = {2003},
NUMBER = {IDSIA-13-03},
ADDRESS = {www.idsia.ch/\-techrep.html},
MONTH = {May},
SOURCE = {OwnPublication},
SOURCETYPE = {TechReport},
ABSTRACT = { One advantage of spiking recurrent neural networks (SNNs)
is an ability to categorise data using a synchrony-based
latching mechnanism. This is particularly useful in
problems where timewarping is encountered, such as speech
recognition. Differentiable recurrent neural networks
(RNNs) by contrast fail at tasks involving difficult
timewarping, despite having sequence learning capabilities
superior to SNNs. In this paper we demonstrate that Long
Short-Term Memory (LSTM) is an RNN capable of robustly
categorizing timewarped speech data, thus combining the
most useful features of both paradigms. We compare its
performance to SNNs on two variants of a spoken digit
identification task, using data from an international
competition. The first task (described in Nature (Nadis
2003)) required the categorisation of spoken digits with
only a single training exemplar, and was specifically
designed to test robustness to timewarping. Here LSTM
performed better than all the SNNs in the competition. The
second task was to predict spoken digits using a larger
training set. Here LSTM greatly outperformed an SNN-like
model found in the literature. These results suggest that
LSTM has a place in domains that require the learning of
large timewarped datasets, such as automatic speech
recognition. }
}
@ARTICLE{eck:psyres2002,
AUTHOR = {D. Eck},
TITLE = {Finding Downbeats with a Relaxation Oscillator},
JOURNAL = {Psychol. Research},
YEAR = {2002},
VOLUME = {66},
NUMBER = {1},
PAGES = {18--25},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2002_psyres.pdf},
SOURCE = {OwnPublication},
SOURCETYPE = {Journal},
ABSTRACT = {A relaxation oscillator model of neural spiking dynamics
is applied to the task of finding downbeats in rhythmical
patterns. The importance of downbeat discovery or {\em beat
induction} is discussed, and the relaxation oscillator
model is compared to other oscillator models. In a set of
computer simulations the model is tested on 35 rhythmical
patterns from Povel \& Essens (1985). The model performs
well, making good predictions in 34 of 35 cases. In an
analysis we identify some shortcomings of the model and
relate model behavior to dynamical properties of relaxation
oscillators. }
}
@ARTICLE{schmidhuber+gers+eck:2002,
AUTHOR = {J. Schmidhuber and F.A. Gers and D. Eck},
TITLE = {Learning Nonregular Languages: A Comparison of Simple
Recurrent Networks and {LSTM}},
JOURNAL = {Neural Computation},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2002_nc.pdf},
YEAR = {2002},
VOLUME = {14},
NUMBER = {9},
PAGES = {2039--2041},
SOURCE = {OwnPublication},
SOURCETYPE = {Journal},
ABSTRACT = {In response to Rodriguez' recent article (Rodriguez 2001)
we compare the performance of simple recurrent nets and
{\em ``Long Short-Term Memory''} (LSTM) recurrent nets on
context-free and context-sensitive languages.}
}
@INPROCEEDINGS{eck+schmidhuber:ieee2002,
AUTHOR = {D. Eck and J. Schmidhuber},
TITLE = {Finding Temporal Structure in Music: Blues Improvisation
with {LSTM} Recurrent Networks},
BOOKTITLE = {Neural Networks for Signal Processing XII, Proceedings of
the 2002 IEEE Workshop},
EDITOR = {H. Bourlard},
PAGES = {747--756},
PUBLISHER = {IEEE},
ADDRESS = {New York},
YEAR = 2002,
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2002_ieee.pdf},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
ABSTRACT = { Few types of signal streams are as ubiquitous as music.
Here we consider the problem of extracting essential
ingredients of music signals, such as well-defined global
temporal structure in the form of nested periodicities (or
{\em meter}). Can we construct an adaptive signal
processing device that learns by example how to generate
new instances of a given musical style? Because recurrent
neural networks can in principle learn the temporal
structure of a signal, they are good candidates for such a
task. Unfortunately, music composed by standard recurrent
neural networks (RNNs) often lacks global coherence. The
reason for this failure seems to be that RNNs cannot keep
track of temporally distant events that indicate global
music structure. Long Short-Term Memory (LSTM) has
succeeded in similar domains where other RNNs have failed,
such as timing \& counting and learning of context
sensitive languages. In the current study we show that LSTM
is also a good mechanism for learning to compose music. We
present experimental results showing that LSTM successfully
learns a form of blues music and is able to compose novel
(and we believe pleasing) melodies in that style.
Remarkably, once the network has found the relevant
structure it does not drift from it: LSTM is able to play
the blues with good timing and proper structure as long as
one is willing to listen. }
}
@UNPUBLISHED{eck:verita2002,
AUTHOR = {D. Eck},
TITLE = {Real Time Beat Induction with Spiking Neurons},
NOTE = {{Music, Motor Control and the Mind: Symposium at Monte
Verita, May}},
YEAR = {2002},
ADDRESS = {},
SOURCE = {OwnPublication},
SOURCETYPE = {Workshop},
ABSTRACT = { Beat induction is best described by analogy to the
activites of hand clapping or foot tapping, and involves
finding important metrical components in an auditory
signal, usually music. Though beat induction is intuitively
easy to understand it is difficult to define and still more
difficult to model. I will discuss an approach to beat
induction that uses a network of spiking neurons to
synchronize with periodic components in a signal at many
timescales. Through a competitive process, groups of
oscillators embodying a particular metrical interpretation
(e.g. \"4/4\") are selected from the network and used to
track the pattern. I will compare this model to other
approaches including a traditional symbolic AI system
(Dixon 2001), and one based on Bayesian statistics (Cemgil
et al, 2001). Finally I will present performance results of
the network on a set of MIDI-recorded piano performances of
Beatles songs collected by the Music, Mind, Machine Group,
NICI, University of Nijmegen (see Cemgil et al, 2001 for
more details or http://www.nici.kun.nl/mmm). }
}
@INPROCEEDINGS{eck+schmidhuber:icann2002,
AUTHOR = {D. Eck and J. Schmidhuber},
TITLE = {Learning The Long-Term Structure of the Blues},
BOOKTITLE = {{Artificial Neural Networks -- ICANN 2002 (Proceedings)}},
EDITOR = {J. Dorronsoro},
VOLUME = {},
PAGES = {284--289},
PUBLISHER = {Springer},
ADDRESS = {Berlin},
YEAR = 2002,
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2002_icannMusic.pdf},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
ABSTRACT = { In general music composed by recurrent neural networks
(RNNs) suffers from a lack of global structure. Though
networks can learn note-by-note transition probabilities
and even reproduce phrases, they have been unable to learn
an entire musical form and use that knowledge to guide
composition. In this study, we describe model details and
present experimental results showing that LSTM successfully
learns a form of blues music and is able to compose novel
(and some listeners believe pleasing) melodies in that
style. Remarkably, once the network has found the relevant
structure it does not drift from it: LSTM is able to play
the blues with good timing and proper structure as long as
one is willing to listen. }
}
@INPROCEEDINGS{gers+perez+eck+schmidhuber:esann2002,
AUTHOR = {F.A. Gers and J.A. Perez-Ortiz and D. Eck and J.
Schmidhuber},
TITLE = {{DEKF-LSTM}},
BOOKTITLE = {Proceedings of the 10th European Symposium on Artificial
Neural Networks, ESANN 2002},
ADDRESS = {},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
YEAR = {2002}
}
@INPROCEEDINGS{gers+perez+eck+schmidhuber:icanna2002,
AUTHOR = {F.A. Gers and J.A. P\'{e}rez-Ortiz and D. Eck and J.
Schmidhuber},
TITLE = {Learning Context Sensitive Languages with {LSTM} Trained
with {Kalman} Filters},
BOOKTITLE = {{Artificial Neural Networks -- ICANN 2002 (Proceedings)}},
EDITOR = {J. Dorronsoro},
PAGES = {655--660},
PUBLISHER = {Springer},
ADDRESS = {Berlin},
YEAR = 2002,
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
ABSTRACT = { Unlike traditional recurrent neural networks, the Long
Short-Term Memory (LSTM) model generalizes well when
presented with training sequences derived from regular and
also simple nonregular languages. Our novel combination of
LSTM and the decoupled extended Kalman filter, however,
learns even faster and generalizes even better, requiring
only the 10 shortest exemplars n <= 10 of the context
sensitive language a^nb^nc^n to deal correctly with values
of n up to 1000 and more. Even when we consider the
relatively high update complexity per timestep, in many
cases the hybrid offers faster learning than LSTM by
itself. }
}
@INPROCEEDINGS{perez+schmidhuber+gers+eck:icannb2002,
AUTHOR = {J.A. P\'{e}rez-Ortiz and J. Schmidhuber and F.A. Gers and
D. Eck},
TITLE = {Improving Long-Term Online Prediction with {Decoupled
Extended Kalman Filters}},
BOOKTITLE = {{Artificial Neural Networks -- ICANN 2002 (Proceedings)}},
EDITOR = {J. Dorronsoro},
PAGES = {1055--1060},
PUBLISHER = {Springer},
ADDRESS = {Berlin},
YEAR = 2002,
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
ABSTRACT = { Long Short-Term Memory (LSTM) recurrent neural networks
(RNNs) outperform traditional RNNs when dealing with
sequences involving not only short-term but also long-term
dependencies. The decoupled extended Kalman filter learning
algorithm (DEKF) works well in online environments and
reduces significantly the number of training steps when
compared to the standard gradient-descent algorithms.
Previous work on LSTM, however, has always used a form of
gradient descent and has not focused on true online
situations. Here we combine LSTM with DEKF and show that
this new hybrid improves upon the original learning
algorithm when applied to online processing. }
}
@TECHREPORT{eck:tr-tracking2002,
AUTHOR = {D. Eck},
TITLE = {Real-Time Musical Beat Induction with Spiking Neural
Networks},
INSTITUTION = {IDSIA},
YEAR = {2002},
NUMBER = {IDSIA-22-02},
ADDRESS = {www.idsia.ch/\-techrep.html},
MONTH = {October},
SOURCE = {OwnPublication},
SOURCETYPE = {TechReport},
ABSTRACT = { Beat induction is best described by analogy to the
activities of hand clapping or foot tapping, and involves
finding important metrical components in an auditory
signal, usually music. Though beat induction is intuitively
easy to understand it is difficult to define and still more
difficult to perform automatically. We will present a model
of beat induction that uses a spiking neural network as the
underlying synchronization mechanism. This approach has
some advantages over existing methods; it runs online,
responds at many levels in the metrical hierarchy, and
produces good results on performed music (Beatles piano
performances encoded as MIDI). In this paper the model is
described in some detail and simulation results are
discussed. }
}
@TECHREPORT{eck:tr-music2002,
AUTHOR = {D. Eck and J. Schmidhuber},
TITLE = {A First Look at Music Composition using {LSTM} Recurrent
Neural Networks},
INSTITUTION = {IDSIA},
YEAR = {2002},
NUMBER = {IDSIA-07-02},
ADDRESS = {www.idsia.ch/\-techrep.html},
MONTH = {March},
SOURCE = {OwnPublication},
SOURCETYPE = {TechReport},
ABSTRACT = { In general music composed by recurrent neural networks
(RNNs) suffers from a lack of global structure. Though
networks can learn note-by-note transition probabilities
and even reproduce phrases, attempts at learning an entire
musical form and using that knowledge to guide composition
have been unsuccessful. The reason for this failure seems
to be that RNNs cannot keep track of temporally distant
events that indicate global music structure. Long
Short-Term Memory (LSTM) has succeeded in similar domains
where other RNNs have failed, such as timing \& counting
and CSL learning. In the current study I show that LSTM is
also a good mechanism for learning to compose music. I
compare this approach to previous attempts, with particular
focus on issues of data representation. I present
experimental results showing that LSTM successfully learns
a form of blues music and is able to compose novel (and I
believe pleasing) melodies in that style. Remarkably, once
the network has found the relevant structure it does not
drift from it: LSTM is able to play the blues with good
timing and proper structure as long as one is willing to
listen.
{\em Note: This is a more complete version of the 2002
ICANN submission Learning the Long-Term Structure of the
Blues.} }
}
@ARTICLE{eck:jnmr2001,
AUTHOR = {D. Eck},
TITLE = {A Positive-Evidence Model for Rhythmical Beat Induction},
JOURNAL = {Journal of New Music Research},
YEAR = {2001},
VOLUME = {30},
NUMBER = {2},
PAGES = {187--200},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2001_jnmr.pdf},
SOURCE = {OwnPublication},
SOURCETYPE = {Journal},
ABSTRACT = {The Normalized Positive (NPOS) model is a rule-based model
that predicts downbeat location and pattern complexity in
rhythmical patterns. Though derived from several existing
models, the NPOS model is particularly effective at making
correct predictions while at the same time having low
complexity. In this paper, the details of the model are
explored and a comparison is made to existing models.
Several datasets are used to examine the complexity
predictions of the model. Special attention is paid to the
model's ability to account for the effects of musical
experience on beat induction.}
}
@INPROCEEDINGS{eck:icann2001,
AUTHOR = {D. Eck},
TITLE = {A Network of Relaxation Oscillators that Finds Downbeats
in Rhythms},
BOOKTITLE = {{Artificial Neural Networks -- ICANN 2001 (Proceedings)}},
EDITOR = {Georg Dorffner},
VOLUME = {},
PAGES = {1239--1247},
PUBLISHER = {Springer},
ADDRESS = {Berlin},
YEAR = 2001,
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2001_icann.pdf},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
ABSTRACT = {A network of relaxation oscillators is used to find
downbeats in rhythmical patterns. In this study, a novel
model is described in detail. Its behavior is tested by
exposing it to patterns having various levels of rhythmic
complexity. We analyze the performance of the model and
relate its success to previous work dealing with fast
synchrony in coupled oscillators. }
}
@INPROCEEDINGS{gers+eck+schmidhuber:icann2001,
AUTHOR = {F. A. Gers and D. Eck and J. Schmidhuber},
TITLE = {Applying {LSTM} to Time Series Predictable Through
Time-Window Approaches},
BOOKTITLE = {{Artificial Neural Networks -- ICANN 2001 (Proceedings)}},
EDITOR = {Georg Dorffner},
PAGES = {669--676},
PUBLISHER = {Springer},
ADDRESS = {Berlin},
YEAR = 2001,
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2001_gers_icann.pdf},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
ABSTRACT = {Long Short-Term Memory (LSTM) is able to solve many time
series tasks unsolvable by feed-forward networks using
fixed size time windows. Here we find that LSTM's
superiority does {\em not} carry over to certain simpler
time series tasks solvable by time window approaches: the
Mackey-Glass series and the Santa Fe FIR laser emission
series (Set A). This suggests t use LSTM only when simpler
traditional approaches fail.}
}
@TECHREPORT{eck:tr-oscnet2001,
AUTHOR = {D. Eck},
TITLE = {A Network of Relaxation Oscillators that Finds Downbeats
in Rhythms},
INSTITUTION = {IDSIA},
YEAR = {2001},
NUMBER = {IDSIA-06-01},
ADDRESS = {www.idsia.ch/\-techrep.html},
MONTH = {February},
PS = {ftp://ftp.idsia.ch/pub/techrep/IDSIA-06-01.ps.gz},
SOURCE = {OwnPublication},
SOURCETYPE = {TechReport},
ABSTRACT = {A network of relaxation oscillators is used to find
downbeats in rhythmical patterns. In this study, a novel
model is described in detail. Its behavior is tested by
exposing it to patterns having various levels of rhythmic
complexity. We analyze the performance of the model and
relate its success to previous work dealing with fast
synchrony in coupled oscillators. \\
{\em Note: See the 2001 ICANN conference proceeding by the
same title for a newer version of this paper.}}
}
@INCOLLECTION{eck+gasser+port:2000,
AUTHOR = {D. Eck and M. Gasser and Robert Port},
TITLE = {Dynamics and Embodiment in Beat Induction},
BOOKTITLE = {{Rhythm Perception and Production}},
EDITOR = {P. Desain and L. Windsor},
PUBLISHER = {Swets and Zeitlinger},
ADDRESS = {Lisse, The Netherlands},
YEAR = 2000,
PAGES = {157--170},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/2000_rppw.pdf},
SOURCE = {OwnPublication},
SOURCETYPE = {Chapter},
ABSTRACT = {We provide an argument for using dynamical systems theory
in the domain of beat induction. We motivate the study of
beat induction and to relate beat induction to the more
general study of human rhythm cognition. In doing so we
compare a dynamical, embodied approach to a symbolic
(traditional AI) one, paying particular attention to how
the modeling approach brings with it tacit assumptions
about what is being modeled. Please note that this is a
philosophy paper about research that was, at the time of
writing, very much in progress. }
}
@TECHREPORT{gers+eck+schmidhuber:tr-2000,
AUTHOR = {F. A. Gers and D. Eck and J. Schmidhuber},
TITLE = {Applying {LSTM} to Time Series Predictable Through
Time-Window Approaches},
INSTITUTION = {IDSIA},
YEAR = {2000},
NUMBER = {IDSIA-22-00},
ADDRESS = {www.idsia.ch/\-techrep.html},
MONTH = {December},
PS = {ftp://ftp.idsia.ch/pub/techrep/IDSIA-22-00.ps.gz},
SOURCE = {OwnPublication},
SOURCETYPE = {TechReport},
ABSTRACT = {Long Short-Term Memory (LSTM) is able to solve many time
series tasks unsolvable by feed-forward networks using
fixed size time windows. Here we find that LSTM's
superiority does {\em not} carry over to certain simpler
time series tasks solvable by time window approaches: the
Mackey-Glass series and the Santa Fe FIR laser emission
series (Set A). This suggests t use LSTM only when simpler
traditional approaches fail.\\
{\em Note: See the 2001 ICANN conference proceeding by the
same title for a newer version of this paper.}}
}
@TECHREPORT{eck:tr-tracking2000,
AUTHOR = {D. Eck},
TITLE = {Tracking Rhythms with a Relaxation Oscillator},
INSTITUTION = {IDSIA},
YEAR = {2000},
NUMBER = {IDSIA-10-00},
ADDRESS = {www.idsia.ch/\-techrep.html},
MONTH = {October},
PS = {ftp://ftp.idsia.ch/pub/techrep/IDSIA-10-00.ps.gz},
SOURCE = {OwnPublication},
SOURCETYPE = {TechReport},
ABSTRACT = { A number of biological and mechanical processes are
typified by a continued slow accrual and fast release of
energy. A nonlinear oscillator exhibiting this slow-fast
behavior is called a relaxation oscillator and is used to
model, for example, human heartbeat pacemaking and neural
action potential. Similar limit cycle oscillators are used
to model a wider range of behaviors including predator-prey
relationships and synchrony in animal populations such as
fireflies. Though nonlinear limit-cycle oscillators have
been successfully applied to beat induction, relaxation
oscillators have received less attention. In this work we
offer a novel and effective relaxation oscillator model of
beat induction. We outline the model in detail and provide
a perturbation analysis of its response to external
stimuli. In a series of simulations we expose the model to
patterns from Experiment 1 of Povel \& Essens (1985). We
then examine the beat assignments of the model. Although
the overall performance of the model is very good, there
are shortcomings. We believe that a network of
mutually-coupled oscillators will address many of these
shortcomings, and we suggest an appropriate course for
future research.\\
{\em Note: See the 2001 {\em Psychological Research}
article "Finding Downbeats with a Relaxation Oscillator"
for a revised but less detailed version of this paper.}}
}
@TECHREPORT{eck:tr-npos2000,
AUTHOR = {D. Eck},
TITLE = {A Positive-Evidence Model for Classifying Rhythmical
Patterns},
INSTITUTION = {IDSIA},
YEAR = {2000},
NUMBER = {IDSIA-09-00},
ADDRESS = {www.idsia.ch/\-techrep.html},
MONTH = {October},
PS = {ftp://ftp.idsia.ch/pub/techrep/IDSIA-09-00.ps.gz},
SOURCE = {OwnPublication},
SOURCETYPE = {TechReport},
ABSTRACT = {The Normalized Positive (NPOS) model is a novel matching
model that predicts downbeat location and pattern
complexity in rhythmical patterns. Though similar models
report success, the NPOS model is particularly effective at
making these predictions while at the same time being
theoretically and mathematically simple. In this paper, the
details of the model are explored and a comparison is made
to existing models. Several datasets are used to examine
the complexity predictions of the model. Special attention
is paid to the model's ability to account for the effects
of musical experience on rhythm perception.\\
{\em Note: See the 2001 Journal of New Music Research paper
"A Positive-Evidence Model for Rhythmical Beat Induction"
for a newer version of this paper.}}
}
@PHDTHESIS{eck:diss,
AUTHOR = {D. Eck},
TITLE = {{Meter Through Synchrony: Processing Rhythmical Patterns
with Relaxation Oscillators}},
SCHOOL = {Indiana University, Bloomington, IN,
www.idsia.ch/\-\~{}doug/\-publications.html },
YEAR = {2000},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/thesis.pdf},
SOURCE = {OwnPublication},
SOURCETYPE = {Thesis},
ABSTRACT = { This dissertation uses a network of relaxation
oscillators to beat along with temporal signals. Relaxation
oscillators exhibit interspersed slow-fast movement and
model a wide array of biological oscillations. The model is
built up gradually: first a single relaxation oscillator is
exposed to rhythms and shown to be good at finding
downbeats in them. Then large networks of oscillators are
mutually coupled in an exploration of their internal
synchronization behavior. It is demonstrated that
appropriate weights on coupling connections cause a network
to form multiple pools of oscillators having stable phase
relationships. This is a promising first step towards
networks that can recreate a rhythmical pattern from
memory. In the full model, a coupled network of relaxation
oscillators is exposed to rhythmical patterns. It is shown
that the network finds downbeats in patterns while
continuing to exhibit good internal stability. A novel
non-dynamical model of downbeat induction called the
Normalized Positive (NP) clock model is proposed, analyzed,
and used to generate comparison predictions for the
oscillator model. The oscillator model compares favorably
to other dynamical approaches to beat induction such as
adaptive oscillators. However, the relaxation oscillator
model takes advantage of intrinsic synchronization
stability to allow the creation of large coupled networks.
This research lays the groundwork for a long-term research
goal, a robotic arm that responds to rhythmical signals by
tapping along. It also opens the door to future work in
connectionist learning of long rhythmical patterns.}
}
@ARTICLE{gasser+eck+port:1999,
AUTHOR = {M. Gasser and D. Eck and R. Port},
TITLE = {Meter as Mechanism: A Neural Network Model that Learns
Metrical patterns},
YEAR = {1999},
JOURNAL = {Connection Science},
VOLUME = {11},
NUMBER = {2},
PAGES = {187--216},
OWN = {Have},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/1999_gasser.pdf},
SOURCE = {OwnPublication},
SOURCETYPE = {Journal},
ABSTRACT = {One kind of prosodic structure that apparently underlies
both music and some examples of speech production is meter.
Yet detailed measurements of the timing of both music and
speech show that the nested periodicities that define
metrical structure can be quite noisy in time. What kind of
system could produce or perceive such variable metrical
timing patterns? And what would it take to be able to store
and reproduce particular metrical patterns from long-term
memory? We have developed a network of coupled oscillators
that both produces and perceives patterns of pulses that
conform to particular meters. In addition, beginning with
an initial state with no biases, it can learn to prefer the
particular meter that it has been previously exposed to.}
}
@INPROCEEDINGS{eck:1999,
AUTHOR = {D. Eck},
TITLE = {Learning Simple Metrical Preferences in a Network of
{F}itzhugh-{N}agumo Oscillators},
BOOKTITLE = {{The Proceedings of the Twenty-First Annual Conference of
the Cognitive Science Society}},
PUBLISHER = {Lawrence Erlbaum Associates},
YEAR = {1999},
EDITOR = {},
ADDRESS = {New Jersey},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
ABSTRACT = {Hebbian learning is used to train a network of oscillators
to prefer periodic signals of pulses over aperiodic
signals. Target signals consisted of metronome-like voltage
pulses with varying amounts of inter-onset noise injected.
(with 0\% noise yielding a periodic signal and more noise
yielding more and more aperiodic signals.) The
oscillators---piecewise-linear approximations (Abbott,
1990) to Fitzhugh-Nagumo oscillators---are trained using
mean phase coherence as an objective function. Before
training a network is shown to readily synchronize with
signals having wide range of noise. After training on a
series of noise-free signals, a network is shown to only
synchronize with signals having little or no noise. This
represents a bias towards periodicity and is explained by
strong positive coupling connections between oscillators
having harmonically-related periods. }
}
@INPROCEEDINGS{chemero+eck:1999,
AUTHOR = {T. Chemero and D. Eck},
TITLE = {An Exploration of Representational Complexity via Coupled
Oscillators},
BOOKTITLE = {{Proceedings of the Tenth Midwest Artificial Intelligence
and Cognitive Science Society}},
YEAR = {1999},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/1999_chemero.pdf},
PUBLISHER = {MIT Press},
ADDRESS = {Cambridge, Mass.},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
ABSTRACT = {We note some inconsistencies in a view of representation
which takes {\it decoupling} to be of key importance. We
explore these inconsistencies using examples of
representational vehicles taken from coupled oscillator
theory and suggest a new way to reconcile {\it coupling}
with {\it absence}. Finally, we tie these views to a
teleological definition of representation.}
}
@INPROCEEDINGS{eck+gasser:1996,
AUTHOR = {D. Eck and M. Gasser},
TITLE = {Perception of Simple Rhythmic Patterns in a Network of
Oscillators},
BOOKTITLE = {{The Proceedings of the Eighteenth Annual Conference of
the Cognitive Science Society}},
PUBLISHER = {Lawrence Erlbaum Associates},
YEAR = {1996},
EDITOR = {},
ADDRESS = {New Jersey},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
ABSTRACT = {This paper is concerned with the complex capacity to
recognize and reproduce rhythmic patterns. While this
capacity has not been well investigated, in broad
qualitative terms it is clear that people can learn to
identify and produce recurring patterns defined in terms of
sequences of beats of varying intensity and rests: the
rhythms behind waltzes, reels, sambas, etc. Our short term
goal is a model which is "hard-wired" with knowledge of a
set of such patterns. Presented with a portion of one of
the patterns or a label for a pattern, the model should
reproduce the pattern and continue to do so when the input
is turned off. Our long-term goal is a model which can
learn to adjust the connection strengths which implement
particular patterns as it is exposed to input patterns.}
}
@INPROCEEDINGS{gasser+eck:1996,
AUTHOR = {M. Gasser and D. Eck},
TITLE = {Representing Rhythmic Patterns in a Network of
Oscillators},
BOOKTITLE = {{The Proceedings of the International Conference on Music
Perception and Cognition}},
PUBLISHER = {Lawrence Erlbaum Associates},
ADDRESS = {New Jersey},
YEAR = {1996},
EDITOR = {},
NUMBER = {4},
PAGES = {361--366},
PDF = {http://www.iro.umontreal.ca/~eckdoug/papers/1996_gasser_icmpc.pdf},
SOURCE = {OwnPublication},
SOURCETYPE = {Conference},
ABSTRACT = { This paper describes an evolving computational model of
the perception and pro-duction of simple rhythmic patterns.
The model consists of a network of oscillators of different
resting frequencies which couple with input patterns and
with each other. Os-cillators whose frequencies match
periodicities in the input tend to become activated.
Metrical structure is represented explicitly in the network
in the form of clusters of os-cillators whose frequencies
and phase angles are constrained to maintain the harmonic
relationships that characterize meter. Rests in rhythmic
patterns are represented by ex-plicit rest oscillators in
the network, which become activated when an expected beat
in the pattern fails to appear. The model makes predictions
about the relative difficulty of patterns and the effect of
deviations from periodicity in the input.}
}
@TECHREPORT{gasser+eck+port:tr-1996,
AUTHOR = {M. Gasser and D. Eck and R. Port},
TITLE = {Meter as Mechanism A Neural Network that Learns Metrical
Patterns},
INSTITUTION = {Indiana University Cognitive Science Program},
YEAR = 1996,
SOURCE = {OwnPublication},
SOURCETYPE = {TechReport},
NUMBER = {180}
}
@MISC{phd:paiement,
AUTHOR = {{J.-F.} Paiement},
TITLE = {A graphical model for music sequence learning},
HOWPUBLISHED = {Ph.D. Dissertation. Ecole Polytechnique F\'{e}d\'{e}rale
de Lausanne (EPFL), Switzerland (Co-supervised with Samy
Bengio)},
YEAR = {2009},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnSupervisedDoc}
}
@MISC{phd:hamel,
AUTHOR = {Philippe Hamel},
TITLE = {Adaptive music generation},
HOWPUBLISHED = {},
YEAR = {to appear},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnSupervisedDoc}
}
@MISC{phd:boulanger-lewandowski,
AUTHOR = {Nicolas Boulanger-Lewandowski},
TITLE = {Timing and dynamics of expressive performance},
HOWPUBLISHED = {},
YEAR = {to appear},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnSupervisedDoc}
}
@MISC{phd:vernays,
AUTHOR = {Michel Vernays},
TITLE = {Computational model of piano timbre},
HOWPUBLISHED = {},
YEAR = {to appear},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnSupervisedDoc}
}
@MISC{ms:bergeron,
AUTHOR = {Arnaud Bergeron},
TITLE = {Automatic expressive performance},
HOWPUBLISHED = {Master's Thesis. University of Montreal Department of
Computer Science and Operations Research},
YEAR = {to appear},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnSupervisedMasters}
}
@MISC{ms:daouda,
AUTHOR = {Tariq Daouda},
TITLE = {Rhythm generation using reservoir computing},
HOWPUBLISHED = {Master's Thesis. University of Montreal Department of
Computer Science and Operations Research (Co-supervised
with Pascal Vincent)},
YEAR = {to appear},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnSupervisedMasters}
}
@MISC{ms:bouchard,
AUTHOR = {Lysiane Bouchard},
TITLE = {Music and machine learning},
HOWPUBLISHED = {Master's Thesis. University of Montreal Department of
Computer Science and Operations Research (Co-supervised
with Pascal Vincent)},
YEAR = {to appear},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnSupervisedMasters}
}
@MISC{ms:lemieux,
AUTHOR = {Simon Lemieux},
TITLE = {Methods for measuring similarity in melodic sequences},
YEAR = {to appear},
HOWPUBLISHED = {Master's Thesis. University of Montreal Department of
Computer Science and Operations Research},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnSupervisedMasters}
}
@MISC{ms:maillet,
AUTHOR = {Francis Maillet},
TITLE = {Automatic mastering of multi-track music recordings using
machine learning},
YEAR = {to appear},
HOWPUBLISHED = {Master's Thesis. University of Montreal Department of
Computer Science and Operations Research},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnSupervisedMasters}
}
@MISC{ms:bertin-mahieux,
AUTHOR = {Thierry Bertin-Mahieux},
TITLE = {Predicting social tags from audio features for music
recommendation},
YEAR = {to appear},
HOWPUBLISHED = {Master's Thesis. University of Montreal Department of
Computer Science and Operations Research (Co-supervised
with Bal\'{a}zs K\'{e}gl)},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnSupervisedMasters}
}
@MISC{ms:wood,
AUTHOR = {Sean Wood},
TITLE = {Using matrix factorization for polyphonic pitch tracking},
YEAR = {to appear},
HOWPUBLISHED = {Master's Thesis. University of Montreal Department of
Computer Science and Operations Research},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnSupervisedMasters}
}
@MISC{ms:lauly,
AUTHOR = {Stanislaus Lauly},
TITLE = {A model of expressive performance timing for the piano},
YEAR = {to appear},
HOWPUBLISHED = {Master's Thesis. University of Montreal Department of
Computer Science and Operations Research},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnSupervisedMasters}
}
@MISC{ms:bergstra,
AUTHOR = {James Bergstra},
TITLE = {Automatic classification of recorded music using machine
learning},
YEAR = {2006},
HOWPUBLISHED = {Master's Thesis. University of Montreal Department of
Computer Science and Operations Research},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnSupervisedMasters}
}
@MISC{ms:lacoste,
AUTHOR = {Alexandre Lacoste},
TITLE = {Machine learning methods for identifying emergent
properties in complex music signals},
YEAR = {2006},
HOWPUBLISHED = {Master's Thesis. University of Montreal Department of
Computer Science and Operations Research},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnSupervisedMasters}
}
@MISC{ms:lapalme,
AUTHOR = {Jasmin Lapalme},
TITLE = {Music composition usion recurrent neural networks and
metrical structure},
YEAR = {2005},
HOWPUBLISHED = {Master's Thesis. University of Montreal Department of
Computer Science and Operations Research},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnSupervisedMasters}
}
@MISC{ms:casagrande,
AUTHOR = {N. Casagrande},
TITLE = {A multi-class, multi-label AdaBoost algorithm that uses
rhythmical and visually-based features for sound and music
classification},
YEAR = {2005},
HOWPUBLISHED = {Master's Thesis. University of Montreal Department of
Computer Science and Operations Research (Co-supervised
with Bal\'{a}zs K\'{e}gl)},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnSupervisedMasters}
}
@MISC{talk:canheit2009,
AUTHOR = {D. Eck},
TITLE = {What can machines learn from music performances?},
HOWPUBLISHED = {The Canadian Higher Education IT Conference, Montreal,
June},
YEAR = {2009},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:vanier2009,
AUTHOR = {D. Eck},
TITLE = {Learning music at multiple timescales},
HOWPUBLISHED = {Vanier College, Montreal, September},
YEAR = {2009},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:ubisoft2009,
AUTHOR = {D. Eck},
TITLE = {Applying Machine Learning to Music and Motion},
HOWPUBLISHED = {Ubisoft Inc., Montreal, January},
YEAR = {2009},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:brams2009,
AUTHOR = {D. Eck},
TITLE = {An Overview of Machine Learning with Applications in Music
and Motion},
HOWPUBLISHED = {BRAMS Scientific Day, Montreal, May},
YEAR = {2009},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:cra2008,
AUTHOR = {D. Eck},
TITLE = {Measuring and Modeling Musical Expression},
HOWPUBLISHED = {CRA Annual Meeting for Computer Science \& Computer
Engineering Chairs, Snowbird, Utah, July},
YEAR = {2008},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:diro2008,
AUTHOR = {D. Eck},
TITLE = {Autotagger: an algorithm for automatically tagging music
collections using large-scale data mining and supervised
machine learning.},
HOWPUBLISHED = {Departmental Colloquium, DIRO, University of Montreal,
Montreal, Canada, May},
YEAR = {2008},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:pop2007,
AUTHOR = {D. Eck},
TITLE = {Panelist},
HOWPUBLISHED = {Pop \& Policy Music Recommendation Panel, Montreal,
Canada, October},
YEAR = {2007},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:google2007,
AUTHOR = {D. Eck},
TITLE = {Automatically Tagging Audio Files Using Supervised
Learning on Acoustic Features},
HOWPUBLISHED = {Google Research Labs, Mountainview Calif, USA, April},
YEAR = {2007},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:sun2006,
AUTHOR = {D. Eck},
TITLE = {An Ensemble Learning Approach to Music Classification},
HOWPUBLISHED = {Sun labs, Sun Microsystems, Boston Mass, USA, April},
YEAR = {2006},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:nyu2006,
AUTHOR = {D. Eck},
TITLE = {Learning Musical Structure},
HOWPUBLISHED = {New York University Faculty of Music, New York, USA,
March},
YEAR = {2006},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:mcgill2005,
AUTHOR = {D. Eck},
TITLE = {Predicting the Similarity Between Music Audio Files},
HOWPUBLISHED = {CS Colloquium, McGill University, Montreal, Canada,
October},
YEAR = {2005},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:cirmmt2005,
AUTHOR = {D. Eck},
TITLE = {Entropy and Autocorrelation: Using Simple Statistics to
Find Tempo and Metrical Structure in Unfiltered Digital
Audio},
HOWPUBLISHED = {CIRMMT, McGill University, Montreal, Canada, March},
YEAR = {2005},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:ofai2005,
AUTHOR = {D. Eck},
TITLE = {Using Autocorrelation to Find Tempo and Structure in
Music},
HOWPUBLISHED = {Austrian Institute for Artificial Intelligence (OFAI),
Vienna, Austria, March},
YEAR = {2005},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:cernec2004,
AUTHOR = {D. Eck},
TITLE = {Methods for Discovering Metrical Structure in Music},
HOWPUBLISHED = {Research Center for Neuropsychology and Cognition
(CERNEC), U. of Montreal, Montreal, Canada, November},
YEAR = {2004},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:ogi2003,
AUTHOR = {D. Eck},
TITLE = {{LSTM} Hybrid Recurrent Networks for Music and Speech},
HOWPUBLISHED = {Oregon Graduate Institute (OGI), Portland, Oregon, USA,
March},
YEAR = {2003},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:losalamos2003,
AUTHOR = {D. Eck},
TITLE = {Timeseries Analysis with {Long Short-Term Memory}
({LSTM})},
HOWPUBLISHED = {Los Alamos National Laboratories (LANL), Los Alamos, New
Mexico, USA, March},
YEAR = {2003},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:ircam2002,
AUTHOR = {D. Eck},
TITLE = {Beat Induction with Spiking Neurons},
HOWPUBLISHED = {L'Institut de Recherche et Coordination Acoustique/Musique
(IRCAM), Paris, France, August},
YEAR = {2002},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:csl2002,
AUTHOR = {D. Eck},
TITLE = {Beat Induction with Spiking Neurons},
HOWPUBLISHED = {Sony Computer Science Laboratory (CSL), Paris, France,
August},
YEAR = {2002},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:usi2002,
AUTHOR = {D. Eck},
TITLE = {Time in Neural Networks},
HOWPUBLISHED = {Università della Svizzera italiana, Lugano, Switzerland,
April},
YEAR = {2002},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:nici2001,
AUTHOR = {D. Eck},
TITLE = {A Nonlinear Dynamical System for Rhythmic Pattern
Discovery},
HOWPUBLISHED = {Nijmegen Institute for Cognition and Information (NICI),
Nijmegen, The Netherlands, July},
YEAR = {2001},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
@MISC{talk:fnn2001,
AUTHOR = {D. Eck},
TITLE = {Finding Temporal Structure in Music using Relaxation
Oscillators},
HOWPUBLISHED = {Foundation for Neural Networks (FNN), Nijmegen, The
Netherlands, July},
YEAR = {2001},
SOURCE = {OwnPublication},
SOURCETYPE = {OwnTalk}
}
This file has been generated by bibtex2html 1.77