LION7 Call for Papers: Special SessionsSpecial sessions are organized as part of LION7 as a way to focus submissions and encourage more interaction between specific communities. In general, submission and publication rules are the same as for the general call for papers, with the organizers of the special sessions coordinating and helping in identifying competent reviewers.
Problem Structure vs. Algorithm Performance in Multiobjective Combinatorial Optimization (LION-PSAP-MOCO)Organizers:
- Hernan Aguirre - Shinshu University, Japan
- Kiyoshi Tanaka - Shinshu University, Japan
- Arnaud Liefooghe - Université Lille 1, France
- Sébastien Verel - Université Nice Sophia-Antipolis, France
Evolutionary algorithms and other classes of metaheuristics are often used to solve difficult problems arising in multiobjective combinatorial optimization. Such randomized search heuristics include evolutionary algorithms, neighborhood-based search, simulated annealing, tabu search, iterated local search, memetic algorithms, hyperheuristics, etc. Successful applications of metaheuristics for multiobjective combinatorial optimization can be found in fields like scheduling, timetabling, planning, network design, transportation and distribution problems, vehicle routing, traveling salesman, packing, power systems, image processing, and many others. However, the relation between the problem structure and the performance of metaheuristics is not yet fully understood. A better understanding would lead to an improved design of algorithms, and an automated setting of parameters (parameter tuning, parameter control).
This special session aims at bringing together researchers working on the design, implementation and experimental analysis of metaheuristics for multiobjective combinatorial optimization. The scope is to identify the main features that make a multiobjective combinatorial optimization problem hard to solve for a metaheuristic, including the search space size, the epistasis, the number of objectives, the objective correlation, or the size and shape of the Pareto front. In addition, we are interested in understanding the behavior and performance of different classes of algorithms with respect to the fitness landscape structure characteristics. Topics of interest include, but are not limited to:
- Multiobjective problem structure analysis
- Algorithm performance
- Algorithm behavior analysis
- Search space analysis, fitness landscapes
- Scalability in the search space (large-scale optimization) and in the objective space (many-objective optimization)
- Theoretical developments
- Classification of multiobjective problem structure / algorithm performance
- Automated tuning and control of parameters
- Neighborhood structures and efficient algorithms for searching them
- Variation operators for evolutionary and other stochastic search methods
- Comparisons between different (also exact) techniques
Multiobjective Optimization and Decision Making (LION-MOO)Organizers:
- Prof. Salvatore Greco, University of Catania, Italy
- Prof. Juergen Branke, University of Warwick, UK
Many practical optimization problems involve multiple objectives. But when one solution is better in one objective, and another solution is better in another objective, the solutions are formally incomparable and decision maker (DM) preferences are needed to make a selection.
In the most common approach, the DM is asked to a priori transform the multi-objective problem into a single-objective problem, e.g., by weighting objectives or turning some objectives into constraints. A different approach is taken by the multi-objective evolutionary algorithm community. Since evolutionary algorithms work with a population of solutions, they can search for an approximation to the entire set of Pareto optimal solutions (solutions that can only be improved in one objective by worsening them in at least one other objective) in a single run. That way, the DM can look at some alternatives and select the most preferred solution a posteriori. In interactive approaches, DM and optimization algorithm interact, and preference information is revealed and refined during optimization.
The aim of a multiobjective optimization procedure should be to guide the decision maker, in an efficient and effective manner, to a preferred solution that is Pareto optimal. The expectation is that an effective learning process would lead to increased satisfaction with and confidence in a decision, as well as a better understanding of the underlying rationale. Therefore, on one hand, a multiobjective optimization procedure should allow the decision maker (DM) to learn about the optimization problem, while, on the other hand, it should be able to learn preferences of the DM to focus the search in the best way. Therefore, we can say that the quality of a multiobjective optimization process is related to what the DM and the algorithm learn.
For this special session, we solicit papers in all areas of multiobjective optimization and decision making, but in particular welcome publications combining learning and optimization.
Dynamic Optimization (LION-DO)Organizers:
- Patrick Siarry, University of Paris-Est Créteil, France
- Amir Nakib , University of Paris-Est Créteil, France
These years dynamic optimization has attracted a growing interest, due to its practical relevance. Indeed, many real-world optimization problems are dynamic, i.e. their objective function changes over time: typical examples are in resource allocation, dynamic vehicle routing, and scheduling, object tracking. In other cases, the objective function is uncertain or noisy as a result of simulation/measurement errors or approximation errors. In addition, the design variables or environmental conditions may also be perturbed or change over time. The objective of an efficient dynamic metaheuristic is to locate the global optimum solution, and to continuously track the optimum in dynamic environments, or to find a robust solution that operates optimally in the presence of uncertainties.
This special session aims at bringing academic researchers and industrials together to review the latest advances and explore future directions in this field. Topics of interest include but are not limited to:
- Benchmark problems and performance measures
- Tracking moving optima
- Dynamic multiobjective optimization
- Adaptation, learning, and anticipation
- Handling noisy fitness functions
- Using fitness approximations
- Searching for robust optimal solutions
- Comparative studies
- Hybrid approaches
- Theoretical analysis
- Real-world applications
In addition, an e-mail message including paper Id, title of paper, authors' names, and abstract must be sent to: nakib[[at]]u-pec.fr and siarry[[at]]u-pec.fr All accepted novel and unpublished papers will be published in the post-conference proceedings of LION7.
Intelligent optimization in Bioinformatics (LION-BIO)Organizers:
- Clarisse Dhaenens, University of Lille 1, INRIA Lille
- Laetitia Jourdan, University of Lille 1, INRIA Lille
Bioinformatics represents a great challenge for optimization methods as many bioinformatics problems can be modelized as large size optimization problems. For example, many bioinformatics problems deal with the manipulation of large sets of variables (SNPs, genes, GWA, proteins ...). Hence, looking for a good combination of these variables require advance search mechanisms. Solving such difficult problems require to incorporate knowledge about problems to be solved.
This special session aims at putting together works in which optimization approaches and knowledge discovery are jointly concerned to solve bioinformatics problems.
Topics of interest include, but are not limited to:
- Original modeling and solving of bioinformatics optimization problems (for example: Folding, docking, protein interaction, network inference etc.)
- Metaheuristics to solve knowledge discovery problems encountered in bioinformatics problems, such as classification, clustering, association rules, feature selection...
- Knowledge discovery approaches embedded in metaheuristics to incorporate knowledge about the problem to be solved
Large Scale Parallelism in Search (LION-PAR)Organizers:
- Pierre Collet, University of Strasbourg, France
- Philippe Codognet, CNRS / UPMC / University of Tokyo, France/Japan
- Florian Richoux, CNRS / University of Tokyo, France/Japan
With the development of multi-core workstations, the availability of GPGPU-enhanced systems, and the access to Grid platforms or supercomputers worldwide, parallel programming is appearing in many domains as a key issue in order to use the computing power at hand in an efficient manner. Heuristic algorithms for hard optimization problems are not isolated from this phenomenon, as bigger computing power means the ability to attack (and solve) more complex problems. Many intelligent optimization algorithms are inherently parallel. In the last decade, various experiments have been done to extend different types of search algorithms (metaheuristics and local search, constraint solvers, SAT solvers, branch & bound) for parallel execution, but most of the time on shared memory multi-core systems (a few cores) or small PC clusters (a few machines or a few tens of machines). The next challenge is thus to devise efficient heuristic search and optimization algorithms for massively parallel computers and heterogeneous systems that will be both scalar and GPU-based and to show their efficiency on very hard optimization problems.
The aim of this special session is to receive papers in the topic of the LION conference (metaheuristics, local search, tabu search, evolutionary algorithms, ant colony optimization, particle swarm optimization, memetic algorithms, and other types of search algorithms) implemented on all kinds of parallel hardware: scalar, GPU-based or heterogeneous massively parallel systems. This workshop is designed to be a forum for researchers willing to tackle those issues, in order to exchange theoretical algorithms and methods, implementation designs, experimental results, to identify future research directions and to further boost this growing area through cross-fertilization.
Papers on all topics related to the session's theme are solicited, in particular: parallelization of existing search algorithms and new parallel methods; heuristic algorithms and combinatorial optimization on Grids, large PC clusters, massively parallel computers, and GPUs; adaptive strategies and learning for parallel search and optimization; applications and benchmarking; theoretical studies and complexity.
Prospective authors can submit papers via the online submission system of LION 7. Authors are advised to be careful when selecting a paper category. The paper categories for LION-Par are (1) LION-Par: Regular Paper, (2) LION-Par: Short paper, and (3) LION-Par: Work for oral presentation only. In addition, an e-mail message including title of paper, author's names and affiliation, and abstract must be sent to the session organizers: pierre.collet [AT] unistra.fr, codognet [AT] is.s.u-tokyo.ac.jp, florian.richoux [AT] polytechnique.edu
All accepted novel and unpublished papers will be published in the post-conference proceedings of LION7.
Games and Computational Intelligence (LION-GCI)Organizers:
- Dr. Antonio J. Fernández Leiva, Universityt of Málaga
- Dr. Antonio M. Mora García, University of Granada
Computational intelligence (CI) comprises a wide set of nature-inspired techniques with an enormous range of practical applications. Problems arising in this area are typically hard and complex to solve to solve, and the associated search spaces are huge. One of the areas that has recently emerged as an exciting field to do research and that provides a high number of interesting problems is the game domain. Games represent fun but also are interesting to study, and provide competitive and dynamic environments that model many real-world problems. On the other hand, CI has been demonstrated to be a powerful tool to be applied in the game domain, including board games, videogames and mathematical games. In addition, CI can make a game more attractive from many points of views, and moreover, games can be used to prove the effectiveness of CI techniques.
This session is aimed to bring together leading researchers and practitioners from academia and industry to discuss recent advances and explore future directions in the synergy between CI and games domains, including the application of CI methods to game domain, or the use of games as platform to value the quality of CI techniques for instance.
The topics of interest include, but are not limited to:
- Bioinspired techniques applied to games (Neural-based systems, Evolutionary algorithms, Ant colony optimization, among others)
- Learning in games
- Coevolution in games
- Fuzzy-based approaches for games
- Artificial intelligent modelling or improvement in games
- Theoretical or empirical analysis of CI techniques for games
- Game theory
- Content generation
- Player satisfaction and experience in games
- Game-based benchmarking
- Computational and artificial intelligence in:
- Board and card games
- Economic or mathematical games
- Serious games
- Augmented and mixed-reality games
- Games for mobile platforms