UAI 2010 Approximate Inference Evaluation – Invitation to participate

Dear Colleagues,

We would like to invite you to participate at the UAI approximate inference challenge, which is now accepting submissions.

Results are published in an online leader-board which is updated regularly (you can use pseudo-names in submissions if you prefer anonymity).

A prize of four free UAI registrations will be awarded to the winning teams, as well as an opportunity to give a short presentation about their
algorithms at the UAI conference in July.

More details are available at: http://www.cs.huji.ac.il/project/UAI10/index.php

The deadline for submissions is July 1st.

Looking forward to your submissions,

Gal Elidan and Amir Globerson

Open Postdoc Positions in Bandits and Reinforcement Learning at INRIA Lille

Open Postdoc Positions in Bandits and Reinforcement Learning at INRIA Lille

The project team SEQUEL (Sequential Learning) of INRIA Lille, France, http://sequel.lille.inria.fr/ is seeking to appoint several Postdoctoral Fellows. We welcome applicants with a strong mathematical background who are interested in theory and applications of reinforcement learning and bandit algorithms.
The research will be conducted under the supervision of Remi Munos, Mohammad Ghavamzadeh and/or Daniil Ryabko, depending on the chosen topics.

The positions are research only and are for one year, with possibility of being extended.
The starting date is flexible, from the Fall 2010 to Spring 2011.

INRIA is France’s leading institution in Computer Science, with over 2800 scientists employed, of which around 250 in Lille. Lille is the capital of the north of France, a metropolis with 1 million inhabitants, with excellent train connection to Brussels (30 min), Paris (1h) and London (1h30).
The Sequel lab is a dynamic lab at INRIA with over 25 researchers (including PhD students) which covers several aspects of machine learning from theory to applications, including statistical learning, reinforcement learning, and sequential learning.

The positions will be funded by the EXPLO-RA project (Exploration-Exploitation for efficient Resource Allocation), a project in collaboration with ENS Ulm (Gilles Stoltz), Ecole des Ponts (Jean Yves Audibert), INRIA team TAO (Olivier Teytaud), Univ. Paris Descartes (Bruno Bouzy), and Univ. Paris Dauphine (Tristan Cazenave).
See: http://sites.google.com/site/anrexplora/ for some of our activities.

Possible topics include:
– In Reinforcement learning: RL in high dimensions. Sparse representations, use of random projections in RL.
– In Bandits: Bandit algorithms in complex environments. Contextual bandits, Bandits with dependent arms, Infinitely many arms bandits. Links between the bandit and other learning problems.
– In hierarchical bandits / Monte-Carlo Tree Search: Analysis and developement of MCTS / hierarchical bandit algorithms, planning with MCTS for solving MDPs
– In Statistical learning: Compressed learning, use of random projections, link with compressed sensing.
– In sequential learning: Sequential prediction of time series

Candidates must have a Ph.D. degree (by the starting date of the position) in machine learning, statistics, or related fields, possibily with background in reinforcement learning, bandits, or optimization.

To apply please send a CV and a proposition of research topic to remi.munos(at)inria.fr or mohammad.ghavamzadeh(at)inria.fr, or daniil.ryabko(at)inria.fr.

If you are planning to go to ICML / COLT this year, we could set up an appointment there.

ICMLA 2010 Speaker Clustering Challenge

Call for Papers

ICMLA 2010 Speaker Clustering Challenge
Washington DC, USA, 12-14 Dec. 2010

http://www.icmla-conference.org/icmla10/CFP_Challenge1_files/CFP_Challenge1.html

OVERVIEW:
Learning methods for sequential data are receiving widespread attention in recent years. This kind of data arises in many interesting scenarios, where the individual semantic units are no longer single vectors but collections of vectors. As examples of these kind of scenarios, we can cite multimedia analysis (e.g., video understanding, speaker recognition), bioinformatics (e.g., DNA or protein sequences), etc. Sequences can have different lengths, so standard distance measures for vector spaces are not directly applicable.
Moreover, sometimes the information conveyed by the sequences is encoded not just on the individual vectors themselves, but also in the dynamics under which these vectors evolve along time. In order to capture such information, it is usual to employ dynamic models such as hidden Markov models or more general dynamic Bayesian networks. Then, distances between sequences can be defined using the learned models.
However, there are many scenarios where the sequences can be accurately classified or clustered without attending their dynamic characteristics. Examples include bag-of-words models for image analysis, speech-independent speaker verification, etc. In these cases the sequences can be viewed as sets of independent and identically distributed (i.i.d.) samples, and can thus be characterized in terms of their underlying probability density function (PDF). There are many ways of defining affinities or distances between PDFs, from the classic Kullback-Leibler or Bhattacharya divergences (even in feature space) to the recently proposed Probability Product Kernels.
In this challenge we propose to focus on unsupervised methods for sequential data. Specifically, clustering of speech data. Clustering tries to find coherent (in some sense) disjoint groups within a dataset. It does not require any training examples, so it is a very important tool for exploratory data analysis. Furthermore, clustering algorithms can be easily expanded into semi-supervised methods which are very useful when the labelling process is costly.

CHALLENGE FORMAT
This challenge proposes two different tasks:
* 2-class speaker clustering
* Multiclass speaker clustering
The first task is 2-class speaker clustering. For this task we provide 7 datasets, each one of them comprised of speech coming from two different speakers. The participants should then identify two clusters within each dataset.
The more advanced task is multiclass speaker clustering. This task is to be carried out on a single dataset, which is formed by sequences coming from an unknown number of speakers in the range. Participants should discover the number of speakers and perform an adequate clustering.
Both tasks are based on a speech database recorded using a PDA. It includes both male and female speakers. Each subject recorded 50 isolated words, and the mean length of each utterance is around 1.3 seconds. The original audio files were processed using the HTK software, yielding a standard parametrization consisting of 12 Mel-frequency cepstral coefficients (MFCCs), an energy term and their respective increments, giving a total of 26 parameters. These parameters were obtained every 10ms with a 25ms analysis window, yielding 26-dimensional sequences of around 130 samples. Any further pre-processing (normalization, filtering, …) is up to the participants.
Participants can submit their results for just one of the tasks or for the two of them. For details on how to format the results, please contact the organizers.

SUBMISSION AND EVALUATION:
Apart from the actual results, a short paper (4 pages) describing the proposed algorithms should be submitted through the main conference submission website. These papers will be reviewed mainly based on:
• Originality and technical soundness of the employed distance measures
• Coherence of the discovered clusters w.r.t. the speakers
• In the multiclass task, special attention will be paid to the steps toward the correct identification of the number of speakers

PUBLICATION:
Accepted papers will be published in the ICMLA’10 conference proceedings.

IMPORTANT DATES:
Paper Submission Deadline: July 15, 2010
Notification of acceptance: September 7, 2010
Camera-ready papers & Pre-registration: October 1, 2010

ICMLA 2010 Challenge Organizers:
* Darío García-García, University Carlos III Madrid, Spain (dggarcia(at)tsc.uc3m.es)
* Raúl Santos-Rodríguez, University Carlos III Madrid, Spain (rsrodriguez(at)tsc.uc3m.es)

CFP: ECCV10 Workshop: Sign Gesture & Activity

International Workshop on Sign Gesture and Activity 2010

Saturday September 11, 2010, Hersonissos, Heraklion, Crete, Greece in conjunction with ECCV 2010.

www.ee.surrey.ac.uk/Personal/R.Bowden/SGA2010

Topic: The workshop will bring together researchers from vision, learning and related areas to present and discuss the recognition of spatio-

temporal motion of people across a broad range of application areas ranging from sign language recognition through to gesture and activity.

Important Dates:

Submission deadlines: Wed, 16th June 2010

Acceptance decisions: Thurs, 8th July 2010

Camera-ready papers: Tue, 13th July 2010

The list of topics will include (but are not limited to):

• Continuous Sign Language Recognition & analysis

• Non-Manual Features and Facial expression recognition

• Feature Extraction for recognition

• Human torso tracking and modelling

• Hand Shape Classification

• Gesture Recognition

• Activity and Action Recognition

• Facial expression analysis

• Lip Reading

• Fusion methods for Recognition

• Multimodal human behaviour analysis

• Non Verbal Communication

• Affective Computing

• Hand and Face Tracking

• Corpora for training and testing

• Semi-automatic corpora annotation tools

• Probabilistic sequence modelling

Invited Speakers: Ivan Laptev, INRIA, France, Dimitris Metaxas, Rutgers, USA

Workshop organisers: Richard Bowden, Uni of Surrey, UK < r.bowden@surrey.ac.uk>

Philippe Dreuw, RWTH Aachen Uni, DE

Petros Maragos, NTUA, Greece,

Justus Piater, Uni of Liège, Belgium,

Submission site: https://cmt.research.microsoft.com/SGA2010

Workshop site: http://www.ee.surrey.ac.uk/Personal/R.Bowden/SGA2010

Main Conference site: http://www.ics.forth.gr/eccv2010/intro.php

Postdoc Positions in New Sheffield Centre

We would like to announce two post-doctoral researcher positions at a
new Sheffield-based Centre for Biosystems Modelling and Inference.
Funded by significant investment from the Faculties of Medicine and
Engineering, , the new center has made three faculty appointments:
Neil Lawrence,
Magnus Rattray and John R. Terry. It is located in a new institute in
a brand new building. The focus of research in the centre will be
probabilistic inference and dynamical modeling.

The two post-doctoral positions are associated with grants
investigating the use of Gaussian process models in biological
systems. The successful candidates will work with Professor Magnus
Rattray and Professor Neil Lawrence on these projects. The
appointments represent an excellent chance to work with a dynamic
group of individuals applying state of the art machine learning
techniques to problems in computational biology.

More details are available here:

Postdoc on the SYNERGY project: http://bit.ly/cLSA9h
Postdoc on Experimental Design: http://bit.ly/duFqN6

Note that the closing date for application for the first position is
soon: 7th July 2010. The second position has a closing date of 23rd
July.

Please contact Magnus or myself if you have any informal queries.

Neil Lawrence
Magnus Rattray

CfA: BCCN 2010 – Berlin Sept 27 – Oct 1

=== Call for Abstracts ===

Bernstein Conference on Computational Neuroscience (BCCN 2010)

The Bernstein Conference on Computational Neuroscience (BCCN) is an
annual meeting of researchers working in Computational Neuroscience
and Neurotechnology. It has grown out of the annual Symposia of the
German National Bernstein Network for Computational Neuroscience,
which have been held since 2005. Now in its 6th year, organized by the
Bernstein Focus: Neurotechnology at the Berlin Institute of
Technology, it has been opened as an international conference. The
BCCN is a single track conference that covers all aspects of
Computational Neuroscience and Neurotechnology. We invite the
submission of abstracts from all relevant areas. Selected abstracts
will be published in the journal Frontiers in Computational
Neuroscience.

The meeting is open for contributions from all relevant areas of
computational neuroscience including, but not limited to: learning and
plasticity, sensory processing, motor control, reward system, brain
computer interface, neural encoding and decoding, decision making,
information processing in neurons and networks, dynamical systems and
recurrent networks, and neurotechnology.

CONFERENCE DATE AND VENUE:
September 27 – October 1, 2010
Technische Universität Berlin
Berlin, Germany

http://www.bccn2010.de/

PHD STUDENT-SYMPOSIUM:
October 1st, 2009
Technische Universität Berlin
Berlin, Germany

IMPORTANT DATES:
Abstract submission deadline: July 2, 2010
Poster submission deadline: July 2, 2010
Notification of acceptance: August 2, 2010
Early registration closes: August 18, 2010

CONFIRMED INVITED SPEAKERS:
Lars-Kai Hansen (Technical University of Denmark)
Ernst Fehr (University of Zurich)
Pascal Fries (Ernst Strüngmann Institute)
Peter Jonas (Albert-Ludwigs-Universität Freiburg)
Misha Tsodyks (Weizmann Institute of Science)
Gero Miesenböck (University of Oxford)

ORGANIZING COMMITTEE:
General Chair: Klaus-Rober Müller
Conference Office: Matthias L. Jugel, Imke Weitkamp

PROGRAM COMMITTEE
Demian Battaglia, Matthias Bethge, Armin Biess, Benjamin Blankertz,
Axel Borst, Martin Burghoff, Gabriel Curio, Ulrich Egert,
Roland Fleming, Alexander Gail, Jan Gläscher, Tim Gollisch,
Ralf Haefner, John-Dylan Haynes, Leo van Hemmen, Andreas Herz,
Frank Hesse, Christian Igel, Dirk Jancke, Christoph Kayser,
Richard Kempter, Peter König, Christian Leibold, Sebastian Möller,
Klaus-Robert Müller, Andreas Neef, Klaus Obermayer, Stefano Panzeri,
Petra Ritter, Constantin Rothkopf, Gregor Schöner, Jens Steinbrink,
Jochen Triesch, Thomas Wachtler, Felix Wichmann, Laurenz Wiskott,
Annette Witt, Gabriel Wittum

Postdoc in Computer Vision and Machine Learning, University of Leeds

Research Fellow in Computer Vision and Machine Learning
University of Leeds – School of Computing

(Full-time, fixed term position for 14 months)

You will work with Dr Mark Everingham (http://www.comp.leeds.ac.uk/me/) on an EPSRC funded project investigating new methods for learning human pose estimation from weak or approximate supervision. The project has three main aims: (i) producing a large dataset of approximately annotated consumer images, at least two orders of magnitude larger than available datasets; (ii) developing machine learning methods to learn from approximate annotation and “side information” for example simple models of human anatomy; (iii) developing strong models of appearance to give robust pose estimation, using the developed machine learning approach. This will include higher order cues modelling appearance of limbs, dependencies between limbs and appearance of joints and configurations of limbs.

You are expected to have a PhD (or to be awarded shortly) in Computer Vision or Machine Learning. You should have experience in developing and applying computer vision and machine learning algorithms, especially probabilistic methods. Expertise in graphical models, structured learning or human pose estimation would be a particular advantage. You should be a proficient programmer in MATLAB and C/C++. You should be self-motivated, good at time management and planning, and have a proven ability to meet deadlines. Good communication and presentation skills are also important.

Salary: Grade 7 (£29,853 – £35,646 p.a)

Apply using: Application form, CV and Equal Opportunities Monitoring form

Application forms:

http://www.leeds.ac.uk/hr/forms/recruitment/app_form.pdf
http://www.leeds.ac.uk/hr/forms/recruitment/newapplicationform.doc

Informal enquiries: Dr Mark Everingham, tel +44 (0)113 343 5370, email m.everingham(at)leeds.ac.uk

Send completed applications to: Judi Drew, email j.a.drew@leeds.ac.uk, or by post to:

Judi Drew
School of Computing
University of Leeds
Leeds
LS2 9JT

Closing date: 18 June 2010

Apply online: http://hr.leeds.ac.uk/jobs/

Phd/Postdoc position in machine learning & signal processing

The Robotics Group of the University of Bremen, Germany, is looking for a highly motivated researcher (PhD student or Postdoc) to work in a project dealing with the online evaluation of EEG data using machine-learning algorithms. Behavioral and fMRI data will also be analyzed during the project. Applicants without a PhD are expected to use their work in order to write a dissertation. The position is planned for 3 years.

The interdisciplinary project cooperates very closely with the German Research Center for Artificial Intelligence (Deutsches Forschungszentrum für künstliche Intelligenz, DFKI). The successful applicant will work together with scientists from many different disciplines (e.g. engineers, computer scientists, physicists, biologists, mathematicians).

Requirements: Applicants must have a university degree (M.Sc. or equivalent) in computer science (or a related field covering the topic, e.g. computational neuroscience or physics) with a strong background in machine learning. Experience with Brain-Computer-Interfaces (BCIs), autonomous mobile robotic systems and/or acquisition of EEG, fMRI or behavioral data is advantageous. Furthermore, experience in programming with C/C++ and/or Python is a benefit.

While English is mandatory, knowledge of German is a plus.

Applications should be sent electronically to verena.tenzer(at)dfki.de using the key number A42/10.

PhD Studentship in Computational Systems Biology (Helsinki, Finland)

PhD studentship in developing novel probabilistic modelling and
statistical inference methodology and applying these methods to
problems in computational systems biology

Department of Information and Computer Science, Aalto University
School of Science and Technology (formerly Helsinki University of
Technology; http://ics.tkk.fi/en/)

The post is funded as part of a large inter-disciplinary European
consortium project on systems approaches to gene regulation biology
through nuclear receptors (SYNERGY) which has been funded under the
ERASysBio+ initiative. The work will involve close collaboration with
other project partners, especially profs. Magnus Rattray and Neil
D. Lawrence (University of Sheffield, UK).

In the project, experimental project partners from Germany and
the Netherlands will generate high-dimensional and multi-modal
time-series datasets interrogating the transcriptional response to
nuclear receptor activation, using an array of next-generation
sequencing experiments. The data will be used to infer and
parameterize systems biology models of cellular function. An iterative
process of model inference and experiment will be used to uncover the
cellular mechanisms of nuclear receptor function and dysfunction. The
methodology developed in the project will build on recent work
integrating probabilistic modelling with systems biology modelling
[e.g. Proc. Natl. Acad. Sci. USA 107(17), 7793 (2010)].

Candidates must have an MSc degree in computer science, electrical
engineering, mathematics, physics, or a related field. A strong
mathematical background and an interest in Bayesian modeling and/or
machine learning. An interest in systems biology is essential but no
prior experience is necessary.

As a PhD student you will join a vibrant research community in the
capital of Finland, characterized by a high quality of life and the
close proximity of nature. The funding is available initially for
three years, with a gross monthly salary ranging from 2000 to 3300
Euro, depending on qualifications and performance. The position will
start on 1 August 2010 or flexibly thereafter. The language of
instruction is English.

For more details and application instructions, see
http://users.ics.tkk.fi/ahonkela/phd_position_2010.shtml

CFP: ECML-SUEMA workshop

ECML – SUEMA 2010 workshop :
Supervised and Unsupervised Ensemble Methods and their Applications
September 20, 2010 – Barcelona, Spain
http://suema10.dsi.unimi.it

Dear Colleague,
we are pleased to invite you to submit a paper to the
workshop Supervised and Unsupervised Ensemble Methods and Their
Applications (SUEMA 2010), organized in the context of the European
Conference on Machine Learning and Principles and Practice of Knowledge
Discovery in Databases (ECML-PKDD 2010).

The workshop is organized with the support of the PASCAL2 (Pattern Analysis,
Statistical Modelling and Computational Learning) European Network of
Excellence.

We are very pleased to announce the PASCAL2 invited speaker,
Grigorios Tsoumakas (Aristotle University of Thessaloniki, Greece);
he will give a talk on “Ensemble Methods for Multi-Label Data”.

SUEMA 2010 intends to provide a forum for researchers in the field
of Machine Learning and Data Mining to discuss topics related to
ensemble methods and their applications.

More information about the topics of the workshop are available at the
workshop web-site: http://suema10.dsi.unimi.it

With best regards

Oleg Okun, Matteo Re and Giorgio Valentini.

— IMPORTANT DATES

Submission 28st June 2010
Notification 19th July 2010
Camera Ready 28st July 2010

— Submission of papers

The authors should submit the papers by e-mail to the workshop chairs
Oleg Okun (olegokun(at)yahoo.com),
Matteo Re (re(at)dsi.unimi.it),
Giorgio Valentini (valentini(at)dsi.unimi.it).

All papers will be peer reviewed based on originality, technical content
and experimental evaluation.

— Workshop Registration

All workshop participants are required to register for the main conference.

— Workshop proceedings

ECML/PKDD will publish all accepted workshop papers on a CD.

As for previous SUEMA editions, workshop chairs are managing to publish
the extended versions of the workshop papers in an edited book or in a
special issue of a machine learning-oriented journal.

—— Main topics

The main topics of the conference include (but are not limited
to):

New ensemble methods raised from new real world supervised and
unsupervised learning problems

Application of ensemble methods in various branches of science
and technology: bioinformatics, medical informatics, computer
security, economics, ecology, meteorology and weather forecast,
image analysis and signal processing, satellite image analysis.

Multi-class, multi-label, multi-path ensemble methods for
hierarchically structured taxonomies.

Fusion of multiple-source/multi-sensor data

Unsupervised ensemble methods for discovering structures in
unlabeled real data

Unsupervised ensemble approaches to assess the
reliability/validity of clusters discovered in real data

Combination techniques and methods to generate multiple base
learners from different features and data

Dynamic member selection for including into an ensemble

Heterogeneous ensembles of base learners

Variants of re-sampling-based methods (bagging, boosting)

Ensemble methods for supervised multi-class classification and
regression

Supervised and unsupervised ensemble methods for structured
domains

Ensemble methods for adaptive incremental learning

— SUEMA Scientific Program Committee

Nicolo’ Cesa-Bianchi, University of Milano, Italy
Carlotta Domeniconi, George Mason University, USA
Robert Duin, Delft University of Technology, the Netherlands
Mark Embrechts, Rensselaer Polytechnic Institute, USA
Ana Fred, Technical University of Lisboa, Portugal
Joao Gama, University of Porto, Portugal
Giorgio Giacinto, University of Cagliari, Italy
Larry Hall, University of South Florida, USA
Ludmila Kuncheva, University of Wales, UK
Francesco Masulli, University of Genova, Italy
Petia Radeva, Autonomous University of Barcelona, Spain
Juan Jose’ Rodriguez, University of Burgos, Spain
Fabio Roli, University of Cagliari, Italy
Paolo Rosso, Polytechnic University Valencia, Spain
Carlo Sansone, Federico II University of Napoli, Italy
Jose’ Salvador Sanchez, University Jaume I, Spain
Grigorios Tsoumakas, Aristotle University of Thessaloniki, Greece
Jordi Vitria’, Autonomous University of Barcelona, Spain
Ioannis Vlahavas, Aristotle University of Thessaloniki, Greece
Terry Windeatt, University of Surrey, UK

— Workshop Chairs

Oleg Okun
Matteo Re
Giorgio Valentini