Fitness landscape Analysis and Search Space Structure

Motivation

Evolutionary algorithms and other classes of meta-heuristics are often used to solve computationally hard optimization problems. Such randomized search heuristics include evolutionary algorithms, neighborhood-based search, simulated annealing, tabu search, iterated local search, memetic algorithms, hyper-heuristics, etc. Successful applications of meta-heuristics can be found in fields like engineering, scheduling, timetabling, planning, network design, transportation and distribution problems, vehicle routing, traveling salesman, packing, power systems, image processing, among many others. This special session aims at bringing together researchers working on understanding the relations between the problem structure and the algorithm performance.

Scope

This session is one of the two sessions dedicated to fitness landscapes analysis. Please visit the web page of the other session: Fitness Landscape Analysis in Practice which is more oriented on practical aspects.

The aim of this special session is to cover rather a more theoretical approach related to fitness landscape analysis, including:

• Understanding problem difficulty and algorithm complexity
• Problem structure analysis, fundamental search space properties
• Gain problem knowledge and learn about the problem structure
• Metrics and descriptors for describing problem features
• Estimating problem features for large-size problem instances
• Characterizing local optimality, ruggedness and neutrality
• Graphs and complex networks for modeling fitness landscapes
• Theory of elementary landscapes
• Algorithm behavior and search performance
• Feature-based performance analysis, correlation and regression
• Choice for representation, evaluation, neighborhood structures and variation operators
• Fitness landscape analysis for the configuration and selection of algorithms
• Automated off-line/tuning and on-line/control/adaptation of parameters, hyper-heuristics
• Performance prediction, performance robustness and scalability issues in the search space (large-scale optimization) and in the objective space (many-objective optimization)
• Software and visualization tools for fitness landscape analysis
• Benchmarking, construction of new models and test problems, problem taxonomies
• Fitness landscape analysis for continuous, combinatorial, mixed and multi-objective optimization

Paper submission

Papers submitted to the special session will be treated as regular papers, and accepted papers will be included in the conference proceedings.

• Papers should be prepared according to the format and page limit of regular papers specified for CEC 2015 (information on the paper format and templates can be found at: http://sites.ieee.org/cec2015/cfp/)
• When submitting your paper, be sure to select the special session name from the main research topic list
• Paper submission should be done through the CEC 2015 website at the following link: http://sites.ieee.org/cec2015/cfp/
top

Important dates

• Paper submission deadline (new extended deadline): January, 16th, 2014
• Paper acceptance notification: February 20, 2015
• Final paper submission deadline: March 13, 2015
• Early registration: March 13, 2015
• Conference dates: May 25-28, 2015

Special session organizers

• Hernan Aguirre, Shinshu University, Japan.
• Arnaud Liefooghe, Université Lille 1, France.
• Kiyoshi Tanaka, Shinshu University, Japan.
• Sébastien Verel, Université du Littoral Côte d'Opale, France.

Programme comittee

(TBC)