The Analytics Department at the IBM Haifa Research Lab (HRL) cordially invites you to a full-day seminar on machine learning, to be held on Sunday, June 7, 2009.

The seminar will take place in the auditorium (room L100) of the HRL site on the Haifa University campus from 09:30 to 18:00. Lunch and light refreshments will be served. Participation is free.

Organizers

• Yishay Mansour, Tel-Aviv University and Google
• Shai Fine, IBM Haifa Research Lab

The goal of Visual Analytics is to derive insight from massive, dynamic, ambiguous, and often conflicting data; detect the expected and discover the unexpected; provide timely, defensible, and understandable assessments; and communicate the assessment effectively for action.

The goal of this workshop is to raise the awareness of the KDD community for the importance of Visual Analytics and bring together researcher from the underlying fields to bridge the gap between them—to write a KDD research roadmap on Visual Analytics. Our invited speakers include Jim Thomas and Daniel A. Keim.*

Function approximation from noisy data is a central task in robot learning. Relevant problems include sensor modeling, manipulation, control, and many others. A large number of function approximation methods have been proposed from statistics, machine learning, and control system theory to address robotics-related issues such as online updates, active sampling, high dimensionality, non-homogeneous noise, and missing features.

In this workshop, we would like to develop a common understanding of the benefits and drawbacks of different function approximation approaches and to derive practical guidelines for selecting a suitable approach to a given problem.

In addition, we would like to discuss two key points of criticism in current robot learning research. First, data-driven machine learning methods do, in fact, not necessarily outperform models designed by human experts and we would like to explore what function approximation problems in robotics really have to be learned. Second, function approximation/regression methods are typically evaluated using different metrics and data sets, making standardized comparisons challenging.

This workshop brings together researches from machine learning, computational finance, academic finance and the financial industry to discuss problems in finance where machine learning may solve challenging problems and provide an edge over existing approaches. The aim of the workshop is to promote discussion on recent progress and challenges as well as on methodological issues and applied research problems. The emphasis will be on practical problem solving involving novel algorithmic approaches.

Topics of the workshop include (but not limited to):

• Optimisation methods
• Reinforcement learning
• Supervised and semi-supervised learning
• Kernel methods
• Bayesian estimation
• Wavelets
• Evolutionary computing
• Recurrent and state space models
• SVMs
• Neural networks
• Boosting
• Multi-agent simulation
• High frequency data
• Trading strategies and hedging techniques
• Execution models
• Forecasting
• Volatility
• Extreme events
• Credit risk
• Portfolio management and optimisation
• Option pricing

Organizers

• David R. Hardoon - University College London
• John Shawe-Taylor - University College London
• Philip Treleaven -University College London
• Laleh Zangeneh - University College London

Pragoramme Committie

• Nicolò Cesa-Bianchi - Università degli Studi di Milano
• Ran El-Yaniv - Technion - Israel Institute of Technology
• Samet Gogus - Barclaycard
• Yuri Kalnishkan - Royal Holloway, University of London
• Jasvindor Kandola - Merrill Lynch
• Donald Lawrence - University College London
• Giuseppe Nuti - Deutsche Bank
• Sandor Szedmak - University of Southampton
• Chris Watkins - Royal Holloway, University of London

Most machine learning (ML) algorithms rely fundamentally on concepts of numerical mathematics. Standard reductions to black-box computational primitives do not usually meet real-world demands and have to be modified at all levels. The increasing complexity of ML problems requires layered approaches, where algorithms are components rather than stand-alone tools fitted individually with much human effort. In this modern context, predictable run-time and numerical stability behavior of algorithms become fundamental. Unfortunately, these aspects are widely ignored today by ML researchers, which limits the applicability of ML algorithms to complex problems.

Our workshop aims to address these shortcomings, by trying to distill a compromise between inadequate black-box reductions and highly involved complete numerical analyses. We will invite speakers with interest in *both* numerical methodology *and* real problems in applications close to machine learning.

While numerical software packages of ML interest will be pointed out, our focus will rather be on how to best bridge the gaps between ML requirements and these computational libraries. A subordinate goal will be to address the role of parallel numerical computation in ML.

Examples of machine learning founded on numerical methods include low level computer vision and image processing, non-Gaussian approximate inference, Gaussian filtering / smoothing, state space models, approximations to kernel methods, and many more.

The workshop will comprise a panel discussion, in which the invited speakers are urged to address the problems stated above, and offer individual views and suggestions for improvement. We highly recommend active or passive attendance at this event. Potential participants are encouraged to contact the organizers beforehand, concerning points they feel should be addressed in this event.

Programme Committee

• Matthias W. Seeger MPI Informatics / Saarland University, Saarbruecken
• Suvrit Sra MPI Biological Cybernetics, Tuebingen
• John P. Cunningham Stanford University (EE), Palo Alto

The main focus of the workshop is the problem of on-line learning when only limited feedback is available to the learner. In on-line learning, at each time step the learner has to predict the outcome corresponding to the next input based on the feedbacks obtained so far. Unlike the usual supervised problem, in which after each prediction the learner is revealed sufficient information to evaluate the goodness of all predictions he could have made, in many cases only limited feedback may be available to the learner. Depending on the nature of the limitation on the feedback, different classes of problems can be identified:

1. Delayed feedback. The utility of an action (i.e., the prediction) is returned only after a certain amount of time. This is the case of reinforcement learning and on-line control problems where the final outcome
of an action may be available only when a goal is finally achieved.

2. Partial feedback. The feedback is limited to that on the learner's prediction so that no information is available on what would other possible predictions bring. Multi-armed bandits, when only the utility of the pulled arm is returned to the learner, is the classic example for this.

3. Indirect feedback. Neither the true outcome, nor the utility of the prediction is observed. Only an indirect feedback loosely related to the prediction is returned.

The increasing interest in on-line learning with limited feedback is also motivated by a number of applications, such as recommender systems, web advertisement systems, in which the user's feedback is limited to accepting/ignoring the proposed item, and the true label (i.e., the item the user would prefer the most) is never revealed to the learner.

Goals

Although some aspects of on-line learning with limited feedback have been already thoroughly analyzed (e.g., multi-armed bandit problems), many problems are still open. For instance, bandits with large action spaces and side information, learning with delayed reward, on-line optimization, etc., are of primary concern in many recent works on on-line learning. Furthermore, on-line learning with limited feedback has strong connections with a number of other fields of Machine Learning such as active learning, semi-supervised learning, and multi-class classification.
The goal of the workshop is to provide researchers with the possibility to present their current research on these topics and to encourage the discussion about the main open issues and the possible connections between the different sub-fields.In particular, we expect the workshop to shed light on a number of theoretical issues, such as:

1. how does the performance of learning algorithms scale in either large (e.g., infinity number of arms, either numerable or continuum, or in metric or measurable spaces) or changing action spaces?
2. how does the performance of learning algorithms scale depending on the smoothness of the function to be optimized (Lipschitz, linear, convex, non convex)?
3. what are the connections between the MDP reinforcement learning paradigm and the on-line learning problem with delayed feedback?
4. how to define complexity measures for on-line learning with limited feedback?
5. is it possible to define a unified view on the problem of learning with delayed, partial, and indirect feedback?

Call for Participation

The organizing committee would like to invite the submission of extended abstracts (three to four pages in the conference format plus appendix if needed) describing research on (but not restricted to) the following topics:

• adversarial/stochastic bandits
• bandits with side information (contextual bandits, associative RL)
• bandits with large and/or changing action spaces
• on-line learning with delayed feedback
• on-line learning in MDPs and beyond
• partial monitoring prediction
• on-line optimization (Lipschitz, linear, convex, non-convex)
• on-line learning in games
• applications

Organisation Committee

• Jean-Yves Audibert (Certis-Université Paris Est-Ecole des Ponts ParisTech)
• Peter Auer (University of Leoben)
• Sebastien Bubeck (INRIA - Team SequeL)
• Alessandro Lazaric (INRIA - Team SequeL) - (primary contact)Odalric Maillard (INRIA - Team SequeL)
• Remi Munos (INRIA - Team SequeL)
• Daniil Ryabko (INRIA - Team SequeL)
• Csaba Szepesvari (University of Alberta)

Applied textual inference has attracted a significant amount of attention in recent years. Recognizing textual entailments and detecting semantic equivalences between texts are at the core of many NLP tasks, including question answering, information extraction, text summarization, and many others. Developing generic algorithms and resources for inference and paraphrasing would therefore be applicable to a broad range of NLP applications.

The success of the first three Recognizing Textual Entailment (RTE) Pascal challenges and the high participation in this year's NIST-organized RTE challenge show that there is a very substantial interest in the area among the research community. RTE and paraphrase detection tasks have considerably stimulated research in the area of applied semantics, and computational models for textual inference are becoming more and more reliable and accurate as a result.

The goal of the workshop is to further stimulate research in these areas, by providing a common forum where people can discuss and compare novel ideas, models and tools for applied textual inference and paraphrasing. The workshop will be open to any research topic related to these area

The aim of the workshop is to discuss various cutting edge design exploration techniques for practical design problems through invited lectures from the machine learning and aerospace disciplines. The workshop is mainly focused on how machine learning methods can be used for aerospace applications, which generally contain multi-objective and multi-disciplinary features.

Workshop Topics:

1. Multi-Objective Design Optimisation
2. Multi-Disciplinary Design Optimisation
3. Robust Design
4. Data Mining
5. Data Fusion

Finally, the workshop aims to encourage researchers to apply their techniques and methods to the presented problems and to encourage continued research into similar problems from the field.

Sparse estimation (or sparse recovery) is playing an increasingly important role in the statistics and machine learning communities. Several methods have recently been developed in both fields, which rely upon the notion of sparsity (e.g. penalty methods like the Lasso, Dantzig selector, etc.). Many of the key theoretical ideas and statistical analysis of the methods have been developed independently, but there is increasing awareness of the potential for cross-fertilization of ideas between statistics and machine learning.

Furthermore, there are interesting links between lasso-type methods and boosting (particularly, LP-boosting); there has been a renewed interest in sparse Bayesian methods. Sparse estimation is also important in unsupervised method (sparse PCA, etc.). Recent machine learning techniques for multi-task learning and collaborative filtering have been proposed which implement sparsity constraints on matrices (rank, structured sparsity, etc.). At the same time, sparsity is playing an important role in various application fields, ranging from image and video reconstruction and compression, to speech classification, text and sound analysis, etc.

The overall goal of the workshop is to bring together machine learning researchers with statisticians working on this timely topic of research, to encourage exchange of ideas between both communities and discuss further developments and theoretical underpinning of the methods.

Organisers

• Sofia Olhede, Department of Statistics, UCL
• Massimiliano Pontil, Department of Computer Science, UCL
• John Shawe-Taylor, Department of Computer Science, UCL

In making advances within Computational Systems Biology there is an acknowledged need for the ongoing development of both probabilistic and mechanistic, possibly multi-scale, models of complex biological processes. In addition to such models the development of appropriate and efficient inferential methodology to identify and reason over such models is necessary.

Examples of the progress which has been made in our understanding of modern biology by the exploitation of such methodology include model based inference of p53 activity; uncovering the evolution of protein complexes and understanding the circadian clock in plants; details of which were presented at the LICSB workshops.

The previous workshop themes of parameter estimation, probabilistic modelling of networks and inference in large biological system models will be further explored in this meeting. Contributions of recent work addressing these themes as well as preliminary application results are welcomed.