<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=ISO-8859-1">
</head>
<body text="#000000" bgcolor="#ffffff">
Call for Papers
<br>
<p>NIPS 2011 WORKSHOP ON MACHINE LEARNING AND INTERPRETATION IN
<br>
NEUROIMAGING
<br>
</p>
<p>(NOTE: this workshop is now MERGED with the NIPS workshop on
<br>
"Interpretable Decoding of Higher Cognitive States from Neural
Data"<br>
- please also note the revised later submission deadline of
October <br>
17th, and that submissions are of 4-page extended abstracts)
<br>
</p>
<p><a target="_blank" rel="nofollow"
href="http://www.google.com/url?sa=D&q=https://sites.google.com/site/mlini2011/&usg=AFQjCNGlTDSGZ2LrKcJVNViB3d2UcdNt0g">https://sites.google.com/site/mlini2011/</a>
<br>
</p>
<p>December 16-17, 2011, Melia Sierra Nevada & Melia Sol y
Nieve, Sierra
<br>
Nevada, Spain
<br>
</p>
<p>Submission deadline (EXTENDED): October 17th, 2011
<br>
</p>
<p>Overview:
<br>
--------------
<br>
</p>
<p>Modern multivariate statistical methods have been increasingly
applied
<br>
to various problems in neuroimaging, including “mind reading”,
“brain
<br>
mapping”, clinical diagnosis and prognosis. Multivariate pattern
<br>
analysis (MVPA) is a promising machine-learning approach for
<br>
discovering complex relationships between high-dimensional signals
<br>
(e.g., brain images) and variables of interest (e.g., external
stimuli
<br>
and/or brain's cognitive states). Modern multivariate
regularization
<br>
approaches can overcome the curse of dimensionality and produce
highly
<br>
predictive models even in high-dimensional, low-sample scenarios
<br>
typical in neuroimaging (e.g., 10 to 100 thousands of voxels and
just
<br>
a few hundreds of samples).
<br>
</p>
<p>However, despite the rapidly growing number of neuroimaging
<br>
applications in machine learning, its impact on how theories of
brain
<br>
function are construed has received little consideration.
Accordingly,
<br>
machine-learning techniques are frequently met with skepticism in
the
<br>
domain of cognitive neuroscience. In this workshop, we intend to
<br>
investigate the implications that follow from adopting machine-
<br>
learning methods for studying brain function. In particular, this
<br>
concerns the question how these methods may be used to represent
<br>
cognitive states, and what ramifications this has for consequent
<br>
theories of cognition. Besides providing a rationale for the use
of
<br>
machine-learning methods in studying brain function, a further
goal of
<br>
this workshop is to identify shortcomings of state-of-the-art
<br>
approaches and initiate research efforts that increase the impact
of
<br>
machine learning on cognitive neuroscience.
<br>
</p>
<p>Decoding higher cognition and interpreting the behavior of
associated
<br>
classifiers can pose unique challenges, as these psychological
states
<br>
are complex, fast-changing and often ill-defined. For instance,
speech
<br>
is received at 3-4 words a second; acoustic, semantic and
syntactic
<br>
processing occur in parallel; and the form of underlying
<br>
representations (sentence structures, conceptual descriptions)
remains
<br>
controversial. ML techniques are required that can take advantage
of
<br>
patterns that are temporally and spatially distributed, but
<br>
coordinated in their activity. And different recording modalities
have
<br>
distinctive advantages: fMRI provides millimeter-level
localization in
<br>
the brain but poor temporal resolution, while EEG and MEG have
<br>
millisecond temporal resolution at the cost of spatial resolution.
<br>
Ideally, machine learning methods would be able to meaningfully
<br>
combine complementary information from these different
neuroimaging
<br>
techniques, and reveal latent dimensions in neural activity, while
<br>
still being capable of disentangling tightly linked and confounded
sub-
<br>
processes.
<br>
</p>
<p>Moreover, from the machine learning perspective, neuroimaging is
a
<br>
rich source of challenging problems that can facilitate
development of
<br>
novel approaches. For example, feature extraction and feature
<br>
selection approaches become particularly important in
neuroimaging,
<br>
since the primary objective is to gain a scientific insight rather
<br>
than simply learn a ``black-box'' predictor. However, unlike some
<br>
other applications where the set features might be quite
well-explored
<br>
and established by now, neuroimaging is a domain where a machine-
<br>
learning researcher cannot simply "ask a domain expert what
features
<br>
should be used", since this is essentially the question the domain
<br>
expert themselves are trying to figure out. While the current
<br>
neuroscientific knowledge can guide the definition of specialized
<br>
'brain areas', more complex patterns of brain activity, such as
spatio-
<br>
temporal patterns, functional network patterns, and other
multivariate
<br>
dependencies remain to be discovered mainly via statistical
analysis.
<br>
</p>
<p>The list of open questions of interest to the workshop includes,
but
<br>
is not limited to the following:
<br>
- How can we interpret results of multivariate models in a
<br>
neuroscientific context?
<br>
- How suitable are MVPA and inference methods for brain mapping?
<br>
- How can we assess the specificity and sensitivity?
<br>
- What is the role of decoding vs. embedded or separate feature
<br>
selection?
<br>
- How can we use these approaches for a flexible and useful
<br>
representation of neuroimaging data?
<br>
- What can we accomplish with generative vs. discriminative
modelling?
<br>
- How can ML techniques help us in modeling higher cognitive
processes
<br>
(e.g. reasoning, communication, knowledge representation)?
<br>
- How can we disentangle confounded processes and representations?
<br>
- How do we combine the data from different recording modalities
<br>
(e.g. fMRI, EEG, structural MRI, DTI, MEG, NIRS, EcOG, single cell
<br>
recordings, etc.)?
<br>
</p>
<p>Workshop Format:
<br>
--------------------------
<br>
</p>
<p>In this two-day workshop we will explore perspectives and novel
<br>
methodology at the interface of Machine Learning, Inference,
<br>
Neuroimaging and Neuroscience. We aim to bring researchers from
<br>
machine learning and neuroscience community together, in order to
<br>
discuss open questions, identify the core points for a number of
the
<br>
controversial issues, and eventually propose approaches to solving
<br>
those issues.
<br>
</p>
<p>The workshop will be structured around 4 main topics:
<br>
- Machine learning and pattern recognition methodology
<br>
- Interpretable decoding of higher cognitive states <br>
- Causal inference in neuroimaging
<br>
- Linking machine learning, neuroimaging and neuroscience
<br>
</p>
<p>Each session will be opened by 2-3 invited talks, and an in depth
<br>
discussion. This will be followed by original contributions.
Original
<br>
contributions will also be presented and discussed during a poster
<br>
session. Each day of the workshop will end with a panel
discussion,
<br>
during which we will address specific questions, and invited
speakers
<br>
will open each segment with a brief presentation of their opinion.
<br>
</p>
<p>This workshop proposal is part of the PASCAL2 Thematic Programme
on
<br>
Cognitive Inference and Neuroimaging (<a class="moz-txt-link-freetext" href="http://">http://</a>
<br>
mlin.kyb.tuebingen.mpg.de/).
<br>
</p>
<p>Paper Submission:
<br>
--------------------------
<br>
</p>
<p>We seek for submission of original (previously unpublished)
research
<br>
papers. The length of the submitted papers should not exceed 4
pages
<br>
in Springer format (here are the LaTeX2e style files), excluding
the
<br>
references. We aim at publishing accepted paper after the workshop
in
<br>
a proceedings volume that contains full papers, together with
short (5-
<br>
page) review papers by the invited speakers. Authors are expected
to
<br>
prepare a full 8 page paper for the final camera ready version
after
<br>
the workshop.
<br>
</p>
<p>Submission of previously published work is possible as well, but
the
<br>
authors are required to mention this explicitly. Previously
published
<br>
work can be presented at the workshop, but will not be included
into
<br>
the workshop proceedings (which are considered peer-reviewed
<br>
publications of novel contributions). Moreover, the authors are
<br>
welcome to present their novel work but choose to opt out of the
<br>
workshop proceedings in case they have alternative publication
<br>
plans.
<br>
</p>
<p>Important dates:
<br>
--------------------------
<br>
</p>
<p>- October 17th, 2011 - paper submission
<br>
- October 24th, 2011 - notification of acceptance/rejection
<br>
- December 16th - 17th - Workshop in Sierra Nevada, Spain,
following
<br>
the NIPS conference
<br>
</p>
<p>Invited Speakers:
<br>
--------------------------
<br>
</p>
<p>Elia Formisano (Universiteit Maastricht, Netherlands)
<br>
Polina Golland (MIT, US)
<br>
James V. Haxby (Dartmouth College, US)
<br>
Tom Mitchell (CMU, US)
<br>
Daniel Rueckert (Imperial College, UK)
<br>
Peter Spirtes (CMU, US)
<br>
Gaël Varoquaux (Neurospin/INRIA, France)
<br>
</p>
<p>Program Committee:
<br>
--------------------------
<br>
Melissa Carroll (Google, New York)
<br>
Guillermo Cecchi (IBM T.J. Watson Research Center)
<br>
Kai-min Kevin Chang, Language Technologies Institute & Centre
for
<br>
Cognitive Brain Imaging, Carnegie Mellon University, Pittsburgh,
USA)
<br>
Moritz Grosse-Wentrup (Max Planck Institute for Intelligent
Systems,
<br>
Tübingen)*
<br>
James V. Haxby (Dartmouth College)
<br>
Georg Langs (Medical University of Vienna)*
<br>
Anna Korhonen (Computer Laboratory & Research Centre for
English and
<br>
Applied Linguistics, University of Cambridge)
<br>
Bjoern Menze (ETH Zuerich, CSAIL, MIT)
<br>
Brian Murphy (Computation, Language and Interaction Group, Centre
for
<br>
Mind/Brain Sciences, University of Trento)*
<br>
Janaina Mourao-Miranda (University College London)
<br>
Vittorio Murino (University of Verona/Istituto Italiano di
Tecnologia)
<br>
Francisco Pereira (Princeton University)
<br>
Irina Rish (IBM T.J. Watson Research Center)*
<br>
Mert Sabuncu (Harvard Medical School)
<br>
Irina Simanova (Max Planck Institute for Psycholinguistics &
Donders
<br>
Institute for Brain, Cognition and Behaviour, Nijmegen)
<br>
Bertrand Thirion (INRIA, NEUROSPIN)
<br>
</p>
<p> Primary contacts:
<br>
</p>
Moritz Grosse-Wentrup <a class="moz-txt-link-abbreviated" href="mailto:moritzgw@ieee.org">moritzgw@ieee.org</a><br>
Georg Langs <a class="moz-txt-link-abbreviated" href="mailto:georg.langs@meduniwien.ac.at">georg.langs@meduniwien.ac.at</a><br>
Brian Murphy <a class="moz-txt-link-abbreviated" href="mailto:brian.murphy@unitn.it">brian.murphy@unitn.it</a>
<br>
Irina Rish <a class="moz-txt-link-abbreviated" href="mailto:rish@us.ibm.com">rish@us.ibm.com</a>
<pre class="moz-signature" cols="72">--
Brian Murphy
Post-Doctoral Researcher
Language, Interaction and Computation Lab
Centre for Mind/Brain Sciences
University of Trento
<a class="moz-txt-link-freetext" href="http://clic.cimec.unitn.it/brian/">http://clic.cimec.unitn.it/brian/</a></pre>
</body>
</html>