5th International Workshop on Machine Learning and Music (MML12)
Research interests
I research the areas of Data Mining, Bioinformatics, and Music
Informatics, within the common theme of sequence data mining. Prior
to my current post I was a Reader in Computer Science at City University London,
and prior to that a Senior Computational Biologist at ZymoGenetics,
Inc., Seattle, mining the human genome for novel potential
biopharmaceuticals.
- Bioinformatics
-
- Sequence analysis. Analysis of next generation sequencing
data, intron/exon boundary detection, SNP discovery, classification
studies in population genetics. Transcriptome assembly algorithms.
Collaboration with the Department of Genetics, University of the Basque Country, on
next generation sequence analysis in the European anchovy
- Invertebrate cytokines.
Collaboration with the Department of Biology, University of Modena and Reggio Emilia, on the
understanding the presence and role of helical cytokines in the
fruitfly Drosophila and other invertebrates such as
Ciona intestinalis
- Music Informatics
-
- Descriptive models. I develop algorithms for pattern
representation and discovery in music, currently with applications to
folk music corpora and music analysis. I collaborate with the
Fundación Euskomedia and
Eresbil archivo vasco de la
música on computational models of Basque folk music.
- Predictive models. Development and evaluation of machine
learning methods for accurate classification of symbolic music. I
research methods that classify music using classifier ensembles, each
modeling some viewpoint of the music. These models are currently
being applied to the task of folk tune classification.
- Chordal analysis. I develop algorithms for annotating
music (symbolic and audio formats) with chord symbols. This is a
challenging problem due to the ambiguity of segment labelling,
combined with the concurrent problem of segmentation into regions to
be labelled.
- Music generation. If we have a good statistical model of
a composer, performer, genre, etc., how can we use that model in an
"inverted" manner to generate music? This broad theme brings up
profound issues in statistical modeling, model architectures, musical
coherence, music knowledge representation, and sampling from
statistical models.
- Music Representation. I develop knowledge representations for
music based on a method called viewpoints, using
methods from description logics and algebraic data types. I'm the maintainer and
developer of the
MIDI-Perl
library.
Research supervision
-
Ruben
Hillewaere, PhD student at Vrije Universiteit Brussel (topic: symbolic music classification)
-
Thomas Rocher, RA (project: Chord Extraction from Audio)
-
Izaro Goienetxea, RA (project: Computational Analysis of Basque Folk Music)