Schedule
The program consists of morning sessions (8:00 - 12:00)
and afternoon sessions (16:00 - 19:00) .
The pause in between (12:00 - 16:00) is planned for informal interaction
and sport activities (the skiing slopes are a short
walking distance from the hotels, as well as skating, snow racket walks, etc.).
The workshop will start with tutorials on Mnday at 0830am. The registration desk shall be open at Hotel Adler (the conferencee venue) from 0600pm on Sunday and 0730am on Monday.
Tutorials and survey talks (MON-TUE-WED)
Lipschitz global optimization and computations with infinite and infinitesimal numbers
Yaroslav Sergeyev (Università della Calabria, Italy)
Abstract
In the first part of the tutorial, Lipschitz global optimization algorithms are presented. One-dimensional geometric and information algorithms are considered. Several approaches to their generalization to the multi-dimensional case are presented. A special attention is paid to techniques used for estimating the Lipschitz constant. The problem of testing of the methods is discussed and the GKLS generator of classes of test functions is presented. The index scheme for solving problems with multiextremal partially defined constrained is described. In the second part of the tutorial, a new way of computations with infinite and infinitesimal quantities is introduced. A new positional system with infinite radix allowing one to write down finite, infinite, and infinitesimal numbers as particular cases of a unique framework is described. This new numeral system gives possibility to introduce a new type of computer able to operate not only with finite numbers but also with infinite and infinitesimal ones; its working simulator is presented. A number of examples showing how the new approach can be used for working with applications (particularly, using functions that can assume infinite and infinitesimal values) is given.Schedule:
- Monday, February 12, 08:30 - 10:30 am
- Monday, February 12, 05:00 - 07:00 pm
Measuring Solution Similarity and Uses for the Measurements
David L. Woodruff (University of California at Davis, USA)
Abstract
In this talk we discuss recent work on measuring the similarity between solutions to hard problems and uses for the measurements. For many kinds of problems there a multiple similarity or distance functions that could reasonably be used; the best one depends on the ultimate application. We will describe examples from knapsack problems, VRPs and general MIPs.Schedule:
- Tuesday, February 13, 10:30 am - 12:30 pm
Elite Pool Maintenance, Fitness Landscape Structure, and Long-Term Memory: Overview, Myths, and Open Frontier
Jean-Paul Watson (Sandia National Laboratories, USA)
Abstract
Pools of high-quality or "elite" solutions are central in the design of advanced memory-based metaheuristics, e.g., those that employ re-intensification or path relinking. The performance of memory-based metaheuristics depends strongly on how the contents of the elite pool are managed over time, and the interaction of long-term memory mechanisms with the pool. Unfortunately, both aspects are poorly understood from a theoretical - or even practical - standpoint, such that the design of pool-based metaheuristics is currently more art than science. This tutorial will survey various approaches to elite pool composition and maintenance in the context of combinatorial metaheuristics, focusing on those aspects and mechanisms currently thought to yield superior performance. The extant theoretical knowledge concerning interaction of elite pools with long-term memory mechanisms will then be discussed. Much of this knowledge is in reality "common wisdom". A large proportion of the tutorial will investigate counter-examples to this common wisdom. The tutorial will conclude with a discussion of what the presenter feels are the most important open research questions in the design and development of pool-based metaheuristics.Schedule:
- Wednesday, February 14, 10:30 am - 12:30 pm
Matheuristics: model-based metaheuristic design
Vittorio Maniezzo (Università di Bologna, Italy)
Abstract
Recently, a community has begun to gather around the idea
that mathematic programming (MP) could be beneficial to metaheuristics.
Currently, metaheuristics (e.g., tabu search, simulated annealing,
genetic algorithms,...) are primal-only methods which have been designed
to be effective, robust and simple. And so they are, usually providing
the best quality solutions on the different benchmark instances reported in
the combinatorial optimization literature. However, there exist problems for
which metaheuristics are not the best performing approach, and there is
anyway the possibility to improve their computational effectiveness both
by integrating MP-based elements into known metaheuristic algorithms and
to design new metaheuristic directly around MP-based concepts. The key idea
is the exploitation of the problem mathematical model in some of the
algorithm steps. A first workshop on this theme was held in summer 2006,
this tutorial will report some of the ideas which were presented.
Schedule:
- Tuesday, February 13, 08:30 - 10:30 am
- Wednesday, February 14, 08:30 - 10:30 am
Reactive search
Roberto Battiti and Mauro Brunato
(Università di Trento, Italy)
Abstract
Most state-of-the-art heuristics are characterized by a certain number of choices
and free parameters, whose appropriate setting is a subject that raises issues of
research methodology. In some cases the role of the user as an intelligent (learning)
part makes the reproducibility
of heuristic results difficult and, as a consequence, the competitiveness of
alternative techniques depends in a crucial way on the user's capabilities.
Reactive Search advocates the use of simple sub-symbolic machine learning to automate
the parameter tuning process and make it an integral (and fully documented)
part of the algorithm. Reactive search deals with modifying some algorithm
parameters in response to the search algorithm's behavior during its execution,
therefore in an online adaptive manner.
If learning is performed on line, task-dependent and
local properties of the configuration space can be used by the algorithm
to determine the appropriate balance between diversification and intensification.
The tutorial presents the basic methods, the software implementation through
appropriate data structures, and relevant applications.
Schedule:
- Monday, February 12, 10:30 am - 12:30 pm
- Tuesday, February 13, 05:00 - 07:00 pm
Technical program (WED-SAT)
(L) = Long talk; (S) = Short talk
Wednesday, February 14
12:30 - 16:30pm: Trip in the snow with snow-shoes / snow-rackets
04:30 - 05:00pm(L): Learning Dynamic Algorithm Portfolios
Matteo Gagliolo and Jürgen Schmidhuber
05:00 - 05:30pm(L): A Hybrid Solver for Constrained Portfolio Selection Problems (preliminary report)
Luca Di Gaspero, Giacomo di Tollo, Andrea Roli, and Andrea Schaerf
05:40 - 06:00pm(S): Reactive and dynamic local search for Max-Clique, an experimental comparison.
Roberto Battiti and Franco Mascia
06:00 - 06:30pm(L): A Continuous-Based Approach for Partial Clique Enumeration
Samuel Rota Bulò, Andrea Torsello, and Marcello Pelillo
06:30 - 07:00pm(L): Evaluating Value Weighting Schemes in the Clustering of Categorical Data
Periklis Andritsos and Vassilios Tzerpos
Thursday, February 15
08:30 - 09:00am(L): Multi-StQPs and their role in Support Vector Machine OptimizationImmanuel M.Bomze, Laura Palagi, and Werner Schachinger (Univ.Vienna)
09:00 - 09:30am(L): Arrhythmia Classification with Confidence-Driven Serial Fusion Based on Support Vector Machines
Aslı U. Özkaya and Fikret S. Gurgen
09:30 - 09:50am(S): Sink Placement in Wireless Sensor Networks Using Genetic Algorithms
Rabun Kosar and Cem Ersoy
10:20 - 10:40am(S): Web Service Orchestration as Constraints Programming
Alexander Lazovik, Marco Aiello, and Rosella Gennari
10:40 - 11:10am(L): Manifold-Based Learning and Search Techniques for Semi-Interactive Global Optimization
Jean-Paul Watson, Brett W. Bader, William M. Brown, Patricia J. Crossno, and Shawn B. Martin
11:10 - 11:30am(S): Machine Learning based histogram matching for Pattern Recognition
Matteo Roffilli, Enrico Angelini, Renato Campanini, and Vittorio Maniezzo
11:30 am - 12:00pm(L): A VNS for MIP: Solving the Connected Facility Location Problem
Ivana Ljubic
04:00 - 05:00pm: Problems & applications Session
Sign Language Recognition: Combining Manual and Non-manual Information
Oya Aran and Lale Akarun
Protein Active Site Detection using SVMs and Kernel Methods (Abstract)
Elisa Cilia
Modelling a Company using Genetic Programming
Juan J. Flores, Mario Graff, and Francisco J. Madrigal
Evolutionary re-scheduling of trains on the French Railway
Yann Semet and Marc Schoenauer
05:15 - 07:00pm: Research and industry panel
Participants:
Davide Bassi (Rector of the University of Trento)
Roberto Battiti (University of Trento)
Luca De Biasi (ArsLogica)
Danilo Di Stefano (ESTECO)
Cem Ersoy (Bogazici University, Turkey)
Holger H. Hoos (University of British Columbia, Canada)
Silvia Poles (ESTECO)
Yaroslav Sergeyev (Università della Calabria, Italy)
Jean-Paul Watson (Sandia National Laboratories, USA)
08:00 pm: Social dinner at:
La tana dell' ermellino.
Friday, February 16
08:30 - 09:00am(L): A Decentralized Strategy for Genetic Scheduling in Heterogeneous EnvironmentsGeorge V. Iordache, Marcela S. Boboila, Florin Pop, Corina Stratan, and Valentin Cristea
09:00 - 09:30am(L): Adaptive traffic flow estimation
R.gabrielli, A.Guidazzi, M.A. Boschetti, V.Maniezzo, and M.Roffilli
09:30 - 10:00am(L): Dynamic Local Search for SAT - Clause Weights, Search Landscapes and Effective Model Finding
Holger Hoos (based on joint work with Dave Tompkins)
10:30 - 11:00am(L): An exact algorithm for Max-Cut
Franz Rendl, Giovanni Rinaldi, and Angelika Wiegele
11:00 - 11:30am(L): Divide-and-Evolve: a New Memetic Scheme for Domain-Independent Temporal Planning
Marc Schoenauer, Pierre Savéant, and Vincent Vidal
11:30am - 12:00pm(L): Dominant Sets and Pairwise Clustering
Massimiliano Pavan and Marcello Pelillo
04:00 - 04:20pm(S): Multilayered Decision Problems
Dmitrii Lozovanu, Stefan Pickl, and Goran Mihelcic
04:20 - 04:40pm(S): GOSH! Gossiping Optimization Search Heuristics
Mauro Brunato, Roberto Battiti, and Alberto Montresor
Invited presentations by local researchers
05:00 - 05:30pm: An Adaptive SVM Nearest Neighbor Classifier for Remotely Sensed ImageryEnrico Blanzieri and Farid Melgani
05:30 - 06:00pm: Cooperative and Competitive Population-Based Approaches for Optimization Problems
M. Donelli, D. Franceschini, and A. Massa
Saturday, February 17
Special session on autonomic systems (CASCADAS project)
This special session is dedicated to the EU-FP6 project CASCADAS.Chair of satellite workshop: Luciano Baresi - Politecnico di Milano
09:00 - 09:30am: A methodology for Model selection by Rare-event simulations and Minimum Cross-entropy principle
Ambedkar Dukkipati and Roberto Battiti - Università di Trento, Italy
09:30 - 10:00am: Selfish node selection for aggregation and load balancing in P2P service networks
Pietro Michiardi - Eurecom, France
10:30 - 11:00am: Self-organisation solutions for knowledge networks and future perspectives
Franco Zambonelli - Università di Modena e Reggio Emilia, Italy
11:00 - 11:30am: Distributed autonomic components and related P2P interaction schemes
Antonietta Mannella - Telecom Italia Lab, Italy
11:30am - 12:00pm: Reputation for self-preservation of autonomic systems
Roberto Cascella - Università di Trento, Italy