CFP: ICML 2011 Workshop on On-line Trading of Exploration and Exploitation 2

ICML 2011 Workshop on On-line Trading of Exploration and Exploitation 2
2 July 2011, Bellevue, Washington, USA

URL: http://explo.cs.ucl.ac.uk/workshop/
Deadline for submission: 28 April 2011

Workshop description
——————————
On-line problems such as website optimisation require to trade exploration and exploitation in order to learn and optimise an unknown target. For instance, in the Pascal Exploration & Exploitation Challenge 2011, a web server observes clicks from visitors to a webpage, and aims to maximise the ratio of clicks per page views. The relationship between the visitor-content pairs and clicks is unknown but it can be learnt from past observations. The content presented to the visitors is either chosen to improve our model of clicks in regions of uncertainty or it is based on the model that has been built so far. These two distinct motives are referred to as exploration and exploitation. Thus, the web server should explore enough to be able to build an accurate model of the visitor’s behaviour, while allowing for sufficient exploitation in order to earn clicks.

The problem of trading exploration and exploitation was first tackled in the “multi-armed bandit” formalism, in which the target observations are compared to rewards obtained when pulling arms of slot-machines. The first theoretical analysis focused on independent reward distributions for each arm. However, in a real-world scenario, arms (inputs) are rarely independent and modelling the dependencies is essential in order to obtain the best learning rate. Gaussian Processes (GP), for instance, are a powerful modelling tool that has been widely used in bayesian online optimisation, in combination with heuristics such as the Most Probable Improvement for selecting inputs where to sample the function. However, performance guarantees for the use of GP in a bandit setting were only found in 2010, when used in combination with the Upper Confidence Bound heuristic for trading exploration and exploitation (Srinivas et al., 2010).

The exploration/exploitation dilemma is a recurrent topic in many areas of research, e.g. global optimisation, reinforcement learning, tree search, recommender systems or information retrieval. The trading of exploration and exploitation is particularly of high importance in various large-scale applications, such as sponsored search advertising (Graepel et al., 2010) or content-based information retrieval (Auer at al., 2010), where the aim is to help users quickly access the information they are looking for. The workshop will provide an opportunity to present, compare and discuss the performance of different exploration/exploitation techniques as well as theoretical analysis of such algorithms. A particular focus of the workshop will be large scale applications. The results of the Exploration and Exploitation Challenge will also be presented during the workshop.

Call for papers
——————————
The workshop will be single-day, comprising of invited talks and presentations of contributed work, with time for discussion. Depending on quality and compatibility with the workshop objectives, slots for brief talks and posters will be allocated.

We invite contributions on the topic of trading exploration and exploitation in various domains including (but not limited to) bandit games, online optimisation, reinforcement learning, tree search, recommender systems, information retrieval, etc. Topics of interest include: applications and practical performance of techniques to trade exploration and exploitation, benchmark/comparison of different techniques, computational challenges in large scale applications, new techniques, theoretical analyses, etc.

Contributions should be communicated to the programme committee (the organisers) in form of an extended abstract (from 4 to 8 pages in the ICML conference paper style), sent by email to: louis.dorard(at)gmail.com .

Important dates
——————————
28 April 2011 – Deadline for abstract submission
20 May 2011 – Notification of Acceptance
2 July 2011 – Workshop taking place in Bellevue, Washington, USA

Workshop organisers
——————————
Louis Dorard, University College London
Suzanne Weller, Adobe
John Shawe-Taylor, University College London
Dorota Glowacka, University College London

Point of contact
——————————
Louis Dorard: louis.dorard(at)gmail.com

The workshop is supported by the Pascal Network of Excellence.

Call for Posters: CVPR’11 Workshop on Fine-Grained Visual Categorization (FGVC)

Deadline for Submission: April 1, 2011 (23:59 GMT)
Notification of Acceptance: Early May, 2011 (4 or 5-week review period)
Workshop date: June 20, 24 or 25, 2011
Website: http://www.fgvc.org

Overview:
Fine categorization, for example the fine distinction into species of
animals and plants, of car and motorcycle models, of architectural
styles, etc., is one of the most interesting and useful open problems
that the machine vision community has yet to confront. Aspects of fine
categorization (also referred to as “subordinate categorization” in
the psychology literature) are discrimination of related categories,
taxonomization, and discriminative vs. generative learning.

Fine categorization lies in the continuum between basic level
categorization (frog vs piano) and identification of individuals (face
recognition, biometrics). The visual distinctions between similar
categories are often quite subtle and therefore difficult to address
with today’s general-purpose object recognition machinery. It is
likely that radical re-thinking of some of the matching and learning
algorithms and models that are currently used for visual recognition
will be needed to approach fine categorization.

This workshop will explore computational questions of modeling,
learning, detection and localization. The invited talks, including
researchers from psychology and psychophysics, will shed light on
human expertise and human performance in subordinate categorization
and taxonomization.

Topics of interest include the following:
– Novel datasets for fine-grained categorization
– Appropriate error metrics for fine-grained categorization
– Constructing field-guides for visual categories
– Embedding human experts’ knowledge into computational Models
– Fine-grained categorization with humans in the loop
– Attribute-based techniques for fine-grained categorization
– Using taxonomies to improve fine-grained categorization
– Part-sharing models for categorization/recognition
– Zero-shot/one-shot recognition
– Transfer-learning from known to novel subcategories
– Domain-specific techniques that generalize to various other domains
– Unsupervised subcategory discovery
– Learning of discriminative features for fine-grained categorization
– Multimodal (e.g. combined audio/video) techniques for fine-grained
categorization

Submission and Reviews:
We invite submission of 1.5-2 page extended abstracts describing work
in the domains suggested above or closely-related areas. Accepted
submissions will be presented as posters (standard CVPR poster format)
at the workshop. Authors are invited to submit a draft or sketch of
the poster’s contents as an optional 3rd page. The extended abstract
and poster sketch should be submitted as a single PDF file with no
more than 3 pages. We also invite authors of accepted CVPR’11 papers
submit an extended abstract of their work to the workshop. Reviewing
of abstract submissions will be carried out by the Program Committee
and will be double-blind.

More information can be found on the FGVC website: http://www.fgvc.org

Symposium on Machine Learning in Speech and Language Processing

June 27, 2011
Bellevue, Washington, USA

http://www.ttic.edu/sigml/symposium2011/

The goal of the symposium is to foster communication and collaboration between researchers
in these synergistic areas, taking advantage of the nearby locations of ACL-HLT 2011 and
ICML 2011. It will bring together members of the Association for Computational
Linguistics, the International Speech Communication Association, and the International
Machine Learning Society.

Topics
=====================

The workshop will feature a series of invited talks and general submissions. Submissions
focusing on novel research are solicited and we especially encourage position and review
papers addressing topics that are relevant both to Speech, Machine Learning and NLP. These
areas include but are not limited to the use of: SVMs, log-linear models, neural networks,
kernel methods, discriminative transforms, large margin training, discriminative training,
active, semi- supervised & unsupervised training, structured prediction, Bayesian modeling,
deep learning , and sparse representations. Application areas include natural language
processing, speech recognition, language modeling, and speaker verification.

Paper Submission
=====================

Prospective authors are invited to submit papers written in English via the “Submissions” link
on the symposium website http://www.ttic.edu/sigml/symposium2011/. Each paper will be
reviewed by at least two reviewers, and each accepted paper must have at least one registered
author.

Confirmed Speakers
=====================

Yoshua Bengio, Jeff Bilmes, Ming-Wei Chang, Stanley Chen, Sanjoy Dasgupta, Mark
Hasegawa-Johnson, Bhuvana Ramabhadran, George Saon, Lawrence Saul, and Mark
Steedman.

Important Dates
=====================

April 15, 2011 Papers due
May 6, 2011 Notification of acceptance
May 27, 2011 Deadline for registration
June 27, 2011 Symposium

Organizing Committee
=====================

Hal Daume III University of Maryland
Joseph Keshet TTI-Chicago
Dan Roth UIUC
Geoffry Zweig Microsoft

Scientific Program Committee
=====================

Jeff Bilmes University of Washington
Brian Kingsbury IBM
Karen Livescu TTI-Chicago

Pascal Exploration & Exploitation Challenge 2011 – Call for submissions

http://explo.cs.ucl.ac.uk/

Are you interested in online learning and website optimisation?

The Exploration & Exploitation Challenge, running under the auspices of the Pascal Network of Excellence and Adobe-Omniture, seeks to improve the relevance of content presented to visitors of a website, based on their individual interests. Participants will submit algorithms whose job will be to predict which visitors are likely to click on which piece of content, and to learn from experience. Submissions will be evaluated on how many clicks they get.

There is information to get started and a screencast at http://explo.cs.ucl.ac.uk/getting-started/ . Please contact us at http://explo.cs.ucl.ac.uk/app/contact if you need any help!

There will also be an ICML workshop on the topic of Exploration & Exploitation, on 2 July 2011, where the final results will be revealed and the winner will be declared. More information is available at http://explo.cs.ucl.ac.uk/workshop/ .

The organisers
Louis Dorard, John Shawe-Taylor, Suzanne Weller, Dorota Glowacka

PhD position in Computational Linguistics and Machine Learning, Saarland University

* Computational Linguistics and Machine Learning
* Saarland University, Saarbruecken, Germany
Cluster of Excellence “Multimodal Computing and Interaction”
* Position open from 1 September 2011
(earlier or later start dates are negotiable)
* Application deadline: 15 April 2011

PROJECT DESCRIPTION

Applications are invited for a PhD position in Computational Linguistics and
Machine Learning. The research will be focused on statistical methods for
inducing structured semantic representation using labelled and unlabelled
data, and different types of prior linguistic knowledge. The goal of this
project is to develop a wide-coverage semantic parser useful for a variety
of natural language processing tasks including question answering,
information extraction and textual entailment. The methods used in this
project will be based on recent advances in non-parametric Bayesian models
for structured prediction and techniques for injecting prior knowledge in
models with latent variables. The prospective student will be co-supervised
by Dr. Caroline Sporleder (http://www.coli.uni-saarland.de/~csporled/)
and Dr. Ivan Titov (http://people.mmci.uni-saarland.de/~titov/).
Additional linguistic expertise will be provided by Dr. Alexis Palmer
(http://comp.ling.utexas.edu/apalmer/).

SAARLAND UNIVERSITY

Saarland University is a European leader in Computer Science research and
teaching, and is especially renowned for its research in Computational
Linguistics and Natural Language Processing. The groups of Caroline
Sporleder and Ivan Titov are a part of the interdisciplinary MMCI Cluster of
Excellence formed in partnership with top-ranked research institutions
located on the Saarland University campus: Max Planck Institute for Computer
Science, Max Planck Institute for Software Systems and German Research
Centre for Artificial Intelligence (DFKI). Researchers come from all over
the world and the research language is English.

The city of Saarbruecken is in the south-west of Germany, close to
France and Luxembourg. Thanks to its proximity to France and its many
students who come from all over the world, it is a very international
city with a lot of French savoir-vivre. Both Paris and Frankfurt can
be reached by train in under two hours. Luxembourg, and the beautiful
French cities of Strasbourg, Nancy, and Metz are also very close by
(1-2 hours by train). The region around Saarbruecken is renowned for
its nature and its mild climate, which makes it an important
wine-growing region. It also boasts several UNESCO World Heritage
Sites (the Voelklinger Huette Ironworks, the Roman-founded city of
Trier, the historic centre of Strasbourg, and the Old Quarters and
Fortifications of Luxembourg).

REQUIREMENTS AND BENEFITS

The ideal candidate for the PhD position should have a masters degree in
computer science or computational linguistics. Some background in machine
learning and computational linguistics is desirable. Minimally, the
candidate should have prior exposure to statistical natural language
processing or strong background in machine learning. The candidate should
also have strong interest in linguistically-motivated statistical models of
language. An excellent academic record, analytical skills and a clear
aptitude for autonomous, creative research will be priority selection
criteria.

The candidates should have strong programming skills as well as excellent
verbal and written communication skills in English.

The salary and social benefits are according to Germany’s public sector TVL
E13 scale, approximately 40,000 Euro p.a. (before taxes), depending on
qualifications and professional experience. The candidate will have the
opportunity of pursuing a doctoral degree at Saarland University.

APPLICATION

The applications should include:

* CV
* brief statement of research interests
* 2-3 references (with email and phone number)
* academic transcript
* list of publications (if any)
* a sample of strongest publications or course work (e.g., Master thesis,
term papers): 1 – 2 papers
Applications (preferably, in a single PDF file) and inquiries should be
directed by email to: Ivan Titov, titov(at)mmci.uni-saarland.de

The position is open until filled. Applications received before April 15
will receive full consideration. The starting date is September 1, 2011 (a
later or earlier starting date is negotiable).

Saarland University wishes to increase the proportion of women in research
and strongly encourages qualified female candidates to apply. Priority will
be given to handicapped candidates with equivalent qualifications.

MORE INFORMATION

Cluster of Excellence MMCI: http://www.mmci.uni-saarland.de/
Saarland University: http://www.uni-saarland.de/en/
The city of Saarbruecken: http://www.saarbruecken.de/en/
The region: http://www.saarland.de/15229.htm

NIPS 2011 Call For Papers

Submissions are solicited for the Twenty-Fifth Annual Conference on Neural
Information Processing Systems, an interdisciplinary conference that brings
together researchers in all aspects of neural and statistical information
processing and computation, and their applications. The conference is a
highly selective, single track meeting that includes invited talks as well
as oral and poster presentations of refereed papers. Submissions by authors
who are new to NIPS are encouraged. In a switch from its previous Vancouver
venue, the 2011 conference will be held on December 13-15 in Granada,
Spain. One day of tutorials (December 12) will precede the main conference,
and two days of workshops (December 16-17) will follow it at the Sierra
Nevada ski resort.

Deadline for Paper Submissions: Thursday June 2, 2011, 23:59 Universal Time
(4:59pm Pacific Daylight Time). Submit at:
https://cmt.research.microsoft.com/NIPS2011/

Technical Areas: Papers are solicited in all areas of neural information
processing and statistical learning, including, but not limited to:

*Algorithms and Architectures: statistical learning algorithms, kernel
methods, graphical models, Gaussian processes, neural networks,
dimensionality reduction and manifold learning, model selection,
combinatorial optimization, relational and structured learning.

*Applications: innovative applications that use machine learning, including
systems for time series prediction, bioinformatics, systems biology,
text/web analysis, multimedia processing, and robotics.

*Brain Imaging: neuroimaging, cognitive neuroscience, EEG
(electroencephalogram), ERP (event related potentials), MEG
(magnetoencephalogram), fMRI (functional magnetic resonance imaging), brain
mapping, brain segmentation, brain computer interfaces.

*Cognitive Science and Artificial Intelligence: theoretical, computational,
or experimental studies of perception, psychophysics, human or animal
learning, memory, reasoning, problem solving, natural language processing,
and neuropsychology.

*Control and Reinforcement Learning: decision and control, exploration,
planning, navigation, Markov decision processes, game playing, multi-agent
coordination, computational models of classical and operant conditioning.

*Hardware Technologies: analog and digital VLSI, neuromorphic engineering,
computational sensors and actuators, microrobotics, bioMEMS, neural
prostheses, photonics, molecular and quantum computing.

*Learning Theory: generalization, regularization and model selection,
Bayesian learning, spaces of functions and kernels, statistical physics of
learning, online learning and competitive analysis, hardness of learning
and approximations, statistical theory, large deviations and asymptotic
analysis, information theory.

*Neuroscience: theoretical and experimental studies of processing and
transmission of information in biological neurons and networks, including
spike train generation, synaptic modulation, plasticity and adaptation.

*Speech and Signal Processing: recognition, coding, synthesis, denoising,
segmentation, source separation, auditory perception, psychoacoustics,
dynamical systems, recurrent networks, language models, dynamic and
temporal models.

*Visual Processing: biological and machine vision, image processing and
coding, segmentation, object detection and recognition, motion detection
and tracking, visual psychophysics, visual scene analysis and
interpretation.

Evaluation Criteria: Submissions will be refereed on the basis of technical
quality, novelty, potential impact, and clarity.

Submission Instructions: All submissions will be made electronically, in
PDF format. As in previous years, reviewing will be double-blind — the
reviewers will not know the identities of the authors. Papers are limited
to eight pages, including figures and tables, in the NIPS style. An
additional ninth page containing only cited references is allowed. Complete
submission and formatting instructions, including style files, are
available from the NIPS website, http://nips.cc.

Supplementary Material: Authors can submit up to 10 MB of material,
containing proofs, audio, images, video, or even data or source code. Note
that the reviewers and the program committee reserve the right to judge the
paper solely on the basis of the 9 pages of the paper; looking at any extra
material is up to the discretion of the reviewers and is not required.

Electronic submissions will be accepted until Thursday June 2, 2011, 23:59
Universal Time (4:59 pm Pacific Daylight Time). As was the case last year,
final papers will be due in advance of the conference.

Dual Submissions Policy: Submissions that are identical (or substantially
similar) to versions that have been previously published, or accepted for
publication, or during the NIPS review period are in submission to another
peer-reviewed and published venue are not appropriate for NIPS, with three
exceptions listed below. These exceptions, which have been approved by the
NIPS Foundation board in the interests of speeding up scientific
communication and improving the efficiency of peer review, are as follows:

1. Concurrent submission to other venues is acceptable provided that: (a) The
concurrent submission or intention to submit to other venues is declared to
all venues, (b) NIPS and the concurrent venues are given permission by the
author(s) to coordinate reviewing, and (c) acceptance to one venue imposes
withdrawal from all other venues with the exception stated in 2 below.

2. NIPS submissions that summarize a longer journal paper, whether published,
accepted, or in submission, are acceptable if the authors inform NIPS and
the journal and give them permission to coordinate reviewing.

3. It is acceptable to submit to NIPS 2011 work that has been made available
as a technical report (or similar, e.g. in arXiv) as long as the conditions
above are satisfied.

None of the above should be construed as overriding the requirements of
other publishing venues. In addition, keep in mind that author anonymity to
NIPS reviewers might be compromised for authors availing themselves of
exceptions 2 and 3.

Authors’ Responsibilities: If there are papers that may appear to violate
any of these conditions, it is the authors’ responsibility to (1) cite
these papers (preserving anonymity), (2) argue in the body of your paper
why your NIPS paper is non-trivially different from these concurrent
submissions, and (3) include anonymized versions of those papers in the
supplemental material.

Demonstrations and Workshops: There is a separate Demonstration track at
NIPS. Authors wishing to submit to the Demonstration track should consult
the Call for Demonstrations. The workshops will be held at the Sierra
Nevada ski resort December 16-17. The upcoming call for workshop proposals
will provide details.

Web URL: http://nips.cc/Conferences/2011/CallForPapers

Scientists piece together EU media structure

PASCAL2 researchers have been involved in a study which analyses how influential the media is in shaping the news agenda in EU Member States.

The researchers, led by Professor Nello Cristianini of University of Bristol, evaluated more than 1 million news articles in 22 languages to identify the factors that make this impact felt. The study, the first mega content analysis of cross-linguistic text using artificial intelligence tools, is presented in the journal PLoS ONE.

More details are available at http://cordis.europa.eu/fetch?CALLER=EN_NEWS_FP7&ACTION=D&DOC=1&CAT=NEWS&QUERY=012ccb4c3b95:66d7:5b8ca464&RCN=32852.

New Book Announcement

Learning with Support Vector Machines

by

Colin Campbell (University of Bristol) and Yiming Ying (University of Exeter)

Published in the Series:
Synthesis Lectures on Artificial Intelligence and Machine Learning
by
Morgan & Claypool Publishers, San Rafael, USA (February 15, 2011)
95 pages
ISBN-10: 1608456161
ISBN-13: 978-1608456161

Support Vectors Machines have become a well established tool within machine learning.
In this book we give a concise overview of this subject. We start with a simple Support
Vector Machine for performing binary classification before considering multi-class
classification and learning in the presence of noise. We show that this framework can
be extended to many other scenarios such as prediction with real-valued outputs,
novelty detection and the handling of complex output structures such as parse trees.
Finally, we give an overview of the main types of kernels which are used in practice and
how to learn and make predictions from multiple types of input data.

Four year post-doctoral position at Royal Holloway, UK

Seeking a talented researcher to do fundamental research on reinforcement learning, working with Chris Watkins. This position is part of the CompLACS project (Composing Learning for Artificial Cognitive Systems).

I am looking for someone who is fascinated by the deep questions of how to model the learning of complex behaviour and the development of intelligence, and who wants to break new ground and develop new approaches in this area. You should also be capable of implementing new algorithms on simulated and real application platforms.

There will be opportunities for collaboration with other consortium members.See here for details of the appointment and how to apply. The deadline is 28th March. I’d be happy to discuss this position informally: my email is chrisw (at) cs.rhul.ac.uk

Deadline 21 March: PhD studentships in Complex and Disordered Systems at King’s College London

Two PhD studentships (Graduate Teaching Assistantships) in Complex and Disordered Systems
King’s College London, Department of Mathematics

The Disordered Systems group at King’s College London expects to have several vacancies for entry into the PhD programme in autumn 2011. The group (see http://www.kcl.ac.uk/schools/nms/maths/research/dissys) has broad-ranging research interests in the application of tools from statistical mechanics to complex and disordered systems, including

– physics: soft matter (phase behaviour and flow), fracture and packing, non-equilibrium and glassy systems
– mathematics: sparse random matrix spectra, localization
– biology: metabolic and protein interaction networks, random graph ensembles, DNA stretching, survival statistics
– econophysics: collective effects in operational risk
– machine learning: learning and statistical inference on graphs

Two PhD studentships have just become available through a Graduate Teaching Assistantship (GTA) scheme, see http://www.kcl.ac.uk/nms/depts/mathematics/people/vacancies.aspx
Successful candidates will be able to gain teaching experience of up to 6 hours per week on average, and receive relevant training in teaching methods. They will receive a studentship payment of around GBP15,000 per year over three years as well as a fee waiver, i.e. they will not have to pay any course fees. This makes the GTAs attractive also to overseas students, for whom course fees can otherwise be substantial.

Interested applicants should contact lucy.ward(at)kcl.ac.uk as soon as possible. The application deadline is very soon, on 21 March 2011. Queries regarding research interests etc can be addressed to Prof Peter Sollich (peter.sollich(at)kcl.ac.uk) or any other member of the research group.