Special sessions have been both a tradition and an essential aspect of IEEE CEC.
With the aim of bringing together researchers on a specific topic, such
sessions are organised by renowned experts in the field across the globe. The
IEEE CEC 2009 Programme Committee solicits proposals for special sessions that
are encompassed within the technical scope of the conference. Papers submitted
for these sessions will be peer-reviewed with the same criteria used for other
contributed papers. All accepted papers in the special sessions will be
included in the published conference proceedings.
Interested researchers are invited to submit a proposal, which should
include the session title, a brief description of the scope and motivation,
names, contact information, and brief CVs of the organisers. It should be made
clear in the application why a special session is needed and how it fits within
the wider scope of the conference. The final submission deadline for proposals
is 1st September 2008, however we are accepting proposals at any
time up to that date.
For enquires, please contact the Special Sessions Chair Dr. Jon Timmis or submit your proposal
via email to Dr Timmis. The accepted special sessions will be posted and
updated here regularly.
Memetic Algorithms (MAs) are one of the recently growing research areas in
Evolutionary Algorithms (EAs). Memetic algorithms are a general name for a
broad class of population-based heuristics that is capable of local
refinements. Recent studies have revealed that MAs are successful on a wide
variety of real world problems. Particularly, they converge to high quality
solutions more efficiently compared to their conventional systematic
counterparts.
A multi-agent system (MAS) is composed of multiple interacting agents, possibly
equipped with intelligent capabilities. By agents here, we typically mean
software agents. Recently, multi-agent systems are increasingly used for
solving problems which are difficult or impossible for an individual agent.
They are also used as a programming and software development paradigm. In a
problem solving multi-agent system, agents usually have some of the basic
properties and characteristics of usual MASs, such as autonomy, local view,
social ability (communications), learning and adaptive ability.
A population in MAs can simply be thought as a collection of agents. In
addition, since MAs are hybrid techniques, as they incorporate both
population-based and local search metaheuristics possibly combined with
tree-search techniques, MASs are indeed a powerful framework for modelling,
designing and implementing them. By integrating the agent concept in MAs, we
can enhance the performance of MAs as evident in the literature. The agents can
bring many interesting features in MAs which are beyond the scope of
traditional evolutionary process and learning.
The aim of this special session is to reflect the most recent advances in the
field, and increase the awareness of the computing community at large on this
effective technology. In particular, we endeavor to demonstrate the current
state-of-the-art in the theory and practice of Agent based MAs. Topics of
interest include (but are not limited to):
Novel frameworks of Agent based MAs (AMAs)
Analytical and/or theoretical studies that enhance our understanding of AMAs
Design of multi-agent architecture within AMAs
Design of agent communication and learning strategy
Analysing the affect of agent type, architecture, cooperation, communication and learning on the overall performance of AMAs
Convergence and complexity analysis of AMAs
AMAs for global, constrained, dynamic and large scale optimization
Biological organisms appear to be orders of magnitude more complex than
present-day computational systems, yet current understanding of their genomes
suggests they consist of only a relatively small number of functional
components. This highlights that the source of biological complexity lies not
within the individual gene products but rather within their interactions and
consequent organisation into biochemical networks.
Biochemical networks are able to carry out very complex behaviours using
relatively few functional components, and they do this using a genetic
representation that is inately evolvable. In essence, this is exactly what is
desired when evolving computation within an evolutionary algorithm: complex
behaviours, a small search space, and an evolvable representation. This has led
to a growing interest in computational models based upon the structure and
behaviour of biochemical networks, such as computational models of genetic
regulatory networks (GRNs), artificial chemistries, and cell signalling models.
These artificial biochemical network models feature a number of related
behaviours, such as self-organisation, feedback, and complex dynamics, and
consequently they face a number of common issues: such as methods of
programming and input-output coupling, the use of appropriate levels of
abstraction, and techniques for analysing, understanding and visualising their
behaviour.
This session aims to bring together researchers interested in artificial
biochemical networks and other computational models motivated by the behaviours
of biological cells. We welcome submissions reporting theoretical or empirical
results for both standalone models and those used within the context of
evolutionary algorithms.
Topics of interest include, but are not limited to:
Computational models of genetic regulatory networks
Algorithmic/Computational chemistries
Computational models of cell signalling
Programmability, scalability, evolvability and fault tolerance
Dynamical systems analysis and mathematical modelling
Implementations and applications, in both hardware and software
Uses within computational development
Cartesian Genetic Programming (CGP)
Session Organisers: James Alfred Walker & Julian Francis Miller
Cartesian Genetic Programming is a form of genetic programming developed
by Julian Miller and Peter Thomson in 1997. In its classic form, CGP
represents a program as a directed graph using a very simple integer-based
representation. In a number of studies, it has been shown to be comparatively
efficient when compared to other GP techniques. Since then, the classical form
of CGP has been enhanced in various ways to include automatically defined
functions (ADFs), multiple chromosomes, and most recently, self-modification
and crossover operators. In addition, CGP has also been applied to a number
of novel and real-world applications in academia and industry.
This is the first special session given on this increasingly popular form of GP.
The aim of this special session is to reflect recent advances and state-of-the-
art developments in CGP, to allow the opportunity for CGP practitioners to
discuss key issues and exchange new ideas regarding CGP, to raise
awareness and the profile of CGP to a wider audience, and to encourage the
growth of the CGP community. The submission of technical and position
papers is invited on all aspects of CGP. Topics of interest include, but are not
limited to:
Applications of CGP to novel and real-world problems
Alternative representations and operators for CGP
Exploiting modularity and problem decomposition in CGP
Novel CGP frameworks
Algorithms using CGP for development and development-based CGP
Analytical, empirical or theoretical analysis that enhances the
Hybridisation of CGP with other evolutionary and bio-inspired
Games are an ideal domain to study computational intelligence methods in
that they provide cheap, competitive, dynamic, reproducible environ ments
suitable for testing new search algorithms, pattern based evaluation methods or
learning concepts. At the same time they are interesting to observe, fun to
play, and very attractive to students.
Computational techniques have successfully been applied to many different
kinds of games, however many research issues are still open. The proposed
session aims at getting together leading researchers and practitioners in this
field who study and apply computational intelligence methods to computer games.
In the context of IEEE CEC 2009 this special session will specifically focus on
those methods that in different ways exploit techniques from the area of
genetic and evolutionary computation, e.g., genetic algorithms, evolutionary
strategies, genetic programming, classifier systems, artificial life,
artificial immune systems, etc. Topics of interest include but they are not
limited to:
Learning and adaptation in games
Knowledge representation in games
Neuro-evolution in games
Coevolution in games
Opponent modelling in games
Knowledge-free and self-learning algorithms in games
Challenges for CI in games
Theoretical or empirical analysis of CI algorithms
Representations for games
Comparative studies (e.g. CI versus human-designed players)
Multi-agent and multi-strategy learning
Board and card games
Economic or mathematical games
Imperfect information and non-deterministic games
Evasion (predator/prey) games
3D computer and console games
"Realistic" games for simulation or training purposes
Games for mobile platforms
Games involving control of physical objects (e.g. remote control car racing)
Games involving physical simulation
Concurrent Approaches to Collaboration in Evolutionary Computation
Session Organisers: Paul Andrews, Adam Sampson, Fiona Polack & Susan Stepney
The study of concurrency encompasses a wide range of techniques including
the use of concurrent programming languages and algorithms and the distribution
of programs across multiple hosts. Concurrency is appealing in nature-inspired
computing for both philosophical and practical reasons. The world around us is
highly concurrent, so concurrent programming techniques provide a natural way
to model entities and interactions in the real world. Evolutionary computing
and bio-inspired agent-based techniques can be very resource intensive
especially for large problems. Concurrency can enable us to harness the huge
computing power available today in multi-core processors, distributed PC
clusters and other parallel systems.
Many bio-inspired computing approaches, such ants and swarms, make use of
interacting and collaborative elements that produce emergent behaviours to
perform their allotted tasks. These approaches are ideally suited to exploit
the power of concurrency to produce performance gains or tackle intensive
problems. Their constituent elements can often be parallelised or distributed,
with communication between elements occurring where necessary. The concurrency
techniques employed to achieve this are often generic, applicable to many
different collaborative evolutionary computing techniques.
We will be interested in receiving both technical and position papers that
make use of concurrent languages, algorithms or distributed techniques in the
following subjects:
Differential evolution, an attractive global optimization method with
relatively fewer parameters, is a relatively new member of the evolutionary
computation family. This method has recently been shown to produce superior
results in a wide variety of real-world applications. This special session
seeks to highlight the latest developments in this rapidly emerging research
area by bringing together researchers and practitioners. Authors are invited to
submit their original and unpublished work to this Special Session. Topics of
interest include, but are not limited to:
Theory of differential evolution
Analysis of parameter settings (scale factor, crossover rate, population size)
Multi-objective differential evolution
Differential evolution for noisy problems
Differential evolution for constrained optimization
Hybridization (with local search and other soft computing approaches)
Connections to / comparison with particle swarm optimization
Connections to other soft computing paradigms (e.g., ANN, Fuzzy systems)
Applications in diverse domains
Evolutionary Algorithms Based on Probabilistic Models
Session Organisers:Qingfu Zhang, José Antonio Lozano, Pedro Larrañaga & Aimin Zhou
Evolutionary algorithms based on probabilistic models (EAPM) have been recognized as a new computing paradigm in
evolutionary computation. Instances of EAPMs include, estimation of distribution algorithms, probabilistic model
building genetic algorithms, ant colony optimization, cross entropy methods, to name a few. There is no traditional
crossover or mutation in EAPMs. Instead, they explicitly extract global statistical information from their previous search
and build a probability distribution model of promising solutions, based on the extracted information. New solutions are
sampled from the model thus built. EAPMs represent a new systematic way to solve hard search and optimization
problems. The last decade has seen growing interest in this area. As an interdisciplinary research area, the development of
EAPMs needs joint efforts from the researchers and practitioners in evolutionary computation, machine learning, statistics
and simulation. This special session aims at bringing researchers who are interested in EAPM together to review the
current state-of-art, exchange the latest ideas and explore future directions. The major topics of interest include, but are
not limited to:
Theory of EAPMs,
New algorithms,
Combination of machine learning techniques and EAPMs,
Both financial and economics problems are more frequently being explore with
Evolutionary Computation (EC) techniques. Theses methods have been proven to be
a powerful tool in domains were analytic solutions are not a good alternative.
Problems in real world involve complexity, noisy environments, imprecision,
uncertainty and vagueness. For this reason EC techniques are needed in order to
solve problems related to these areas. So far it has been successfully used for
estimating econometric parameters, macroecomics models, replicating laboratory
results with human subjects, searching equilibriums, studying the emergence of
the representative agent and rational expectations, designing public policy, in
financial engineering, risk management, portfolio optimization, industrial
organization, auctions, experimental economics, financial forecasting, market
simulation or agent-based computational economics among many other areas.
Evolutionary Computation in Bioinformatics and Computational Biology
Bioinformatics and computational biology present a number of difficult
optimization problems with large search spaces. Recent applications of
evolutionary computation in this area suggest that they are well-suited to this
area of research. This special session will highlight applications of
evolutionary computation to a broad range of topics. Particular interest will
be directed towards novel applications of evolutionary computation to problems
in these areas.
The bioinformatics special session has been a part of CEC since 1999 and is
soliciting high quality papers of original research and application papers that
have not been published elsewhere and are not under consideration for
publication elsewhere. All papers will be rigorously reviewed by at least 2
reviewers. Accepted papers will be published in the CEC proceedings. There
is a clear interest in both the computational intelligence comunity and biology
communities for this special session.
Evolutionary Computation in Dynamic and Uncertain Environments
Session Organisers: Shengxiang Yang, Hans-Georg Beyer, Yaochu Jin & Ponnuthurai N. Suganthan
Many real-world optimization problems are subjected to dynamic and uncertain
environments that are often impossible to avoid in practice. For instance, the
fitness function is uncertain or noisy as a result of simulation/measurement
errors or approximation errors (in the case where surrogates are used in place
of the computationally expensive high fidelity fitness function). In addition,
the design variables or environmental conditions many also perturb or change
over time. For these dynamic and uncertain optimization problems the objective
of the evolutionary algorithm is no longer to simply locate the global optimum
solution, but to continuously track the optimum in dynamic environments, or to
find a robust solution that operates optimally in the presence of
uncertainties. This poses serious challenges to conventional evolutionary
algorithms.
Evolutionary Computation in Network-on-Chip Based Systems
Session Organisers: Nadia Nedjah & Luiza de Macedo Mourelle
Network-on-Chip (NoC) is an emerging paradigm for communications within
large VLSI systems implemented on a single silicon chip. IT is used as a new
approach to designing complex System-on-a-chip (SoCs) design. NoC-based
systems can accommodate multiple complex SoC designs. In a NoC-based system,
modules such as processor cores, memories and specialized IP blocks exchange
data using a on-chip network. An NoC is constructed from multiple
point-to-point data links interconnected by switches also called routers, such
that messages can be relayed from any source module to any destination module
over several links, by making routing decisions at the switches.
VLSI designers of NoC-based systems face several problems, among which we
can cite, for instance, planning the architecture that is most suitable to a
given application is order to improve performance and mapping the sub- systems
that form the application into a multiple nodes of the NoC architecture.
Evolutionary computation can be used as a very robust tool to bring some
answers to this kind of design problems.
The aim of this special session is to bring together hardware, middleware
and application designers that exploit the evolutionary computation principles
to provide CAD tools for NoC-based systems. This session will allow researchers
to share experiences and identify theoretical and technical issues in this
field of expertise. Submitted papers may describe applications, computing
models, modeling frameworks, or hardware platforms and architecture.
Evolutionary Computation in Scheduling and Planning
Session Organisers: Lam T. Bui, James M. Whitacre & Hussein A. Abbass
Scheduling and planning problems are generally complex, constrained and
multi-objective. The application of evolution-inspired, meta-heuristic and
other soft computing techniques to this problem domain has received
considerable attention from the research community.
This special session on Evolutionary Computation in Scheduling and Planning
(ECSP) seeks to bring together researchers from around the globe for a creative
discussion on recent advances and challenges facing ECSP research.
Evolutionary Computation in Space and Air
Session Organisers: Massimiliano Vasile, Oliver Schuetze, David W. Corne & Bruce Conway
In the Aeropsace Sciences, many applications require the solution of global
single and multi-objective optimization problems or problems with mixed
variables and non-differentiable quantities. From global trajectory
optimization to multidisciplinary aircraft and spacecraft design, from planning
and scheduling for autonomous vehicles to the synthesis of robust controllers
for airplanes or satellites, evolutionary based techniques have become an
important tool for tackling these kinds of problems providing interesting
solutions. Not only has this given the way to application of evolutionary
computation but has led also to the development of new approaches.
In most of the cases the basic evolutionary heuristics have been hybridized
with other techniques, such as gradient methods or branch and prune methods, or
modified to better adapt to the specific application under investigation. This
has led to the creation of new heuristics, new meta-heuristics or new
hybridizations that have proven to be very effective.
Evolutionary Computer Vision
Session Organisers: Vic Ciesielski, Mario Koeppen & Mengjie Zhang
Computer vision is a major unsolved problem in computer science and
engineering. Over the last decade there has been increasing interest in using
evolutionary computation approaches to solve vision problems. Computer vision
provides a range of problems of varying difficulty for the development and
testing of evolutionary algorithms.
The theme of the proposed special session is the use of evolutionary
computation for solving computer vision and image processing problems. This
special session seeks to highlight the latest developments in this research
area by bringing together researchers and practitioners in both evolutionary
computation and computer vision. Authors are invited to submit their original
and unpublished work to this Special Session.
Evolutionary Development
Session Organisers: Till Steiner, Gunnar Tufte & Markus Olhofer
This special session of the IEEE 2009 CEC Congress on Evolutionary
Computation will focus on the design and analysis of developmental systems in
evolutionary computation.
Over the recent years researchers in the evolutionary computation comunity
have created an increasing number of evolutionary developmental systems with
varying levels of complexity. Much attention has been paid to the creation of
these systems and the evaluation of their abilities to produce large, complex,
modular, and robust phenotypes.
Due to the inherent complexity of developmental systems and of the created
solutions, the analysis of developmental processes and their outcome proves to
be very difficult. Results often have to be restricted to basic experimental
status, whilst a detailed understanding of the dynamics in the system is
frequently not available. To address these issues and to lead science towards
an enhanced understanding of the processes involved in artificial EvoDevo
systems, a revision of methods and tools for analysis seems necessary.
Concepts and ideas for the development of these methods might be found by
simultaneously integrating multiple scales, combining for example the dynamics
on developmental timescale with dynamics on evolutionary timescale, or for
multi-cellular representations, the behaviour of single cells, groups of cells,
and the resulting character of the phenotype.
The aim of this special session is to promote discussion of evolutionary
developmental systems with a focus on their analysis and understanding, as well
as to suggest possible approaches to exploit the features unique to
developmental systems with respect to system design.
Topics of interest include but are not limited to:
analysis and modeling of dynamic systems for genotype-phenotype mapping
generative systems
analysis and modeling of regulatory networks for evolutionary development
multi-scale and multi-level modeling of evolutionary development
regulatory vs. functional mechanisms in evolutionary development
applications for evolutionary development
transfer of system-level biological properties to computational and technical systems
relation between structural and functional analysis of biological systems
The aim of this session is to bring together researchers working in the area of evolutionary games
with relation to network effects.
The emergence of cooperation in overcoming a dilemma can be explained by several theories such
as kin selection, direct reciprocity, indirect reciprocity, network reciprocity, and group selection.
Network reciprocity is one of the most important ideas among them, since the network reciprocity
can make altruism emerge, even though requiring that agents use only the simplest strategy-either
cooperation (C) or defection (D). Thus, the network reciprocity may explain why a number of
animal species, unsophisticated in terms of information processing, have evolved cooperative social
systems. Observing ourselves, the network reciprocity might be able to give a plausible answer of
why human society evolves complex networks to solve conflicting problems that are often
accompanied by a heavy social dilemma. Thus, evolutionary games on complex networks shed a
clear light on those unsolved inquiries in evolutionary biology, sociobiology and other social
sciences. Moreover, recently, the evolutionary games on complex networks also call particular
interests from interdisciplinary areas of nonlinear science, because physicists have observed that
there are several analogies between emergence of cooperation on evolutionary network games and
phase transition of crystal lattice structures.
Evolutionary Robotics
Session Organisers: Patricia A. Vargas, Sabine Hauert, Dario Floreano & Phil Husbands
Evolutionary Robotics (ER) aims to apply evolutionary computation
techniques, inspired by Darwin's principle of selective reproduction of the
fittest, to automatically design the control and/or hardware of both real and
simulated autonomous robots.
Having an intrinsic interdisciplinary character, ER is being employed
towards the development of many fields of research, among which we can
highlight neuroscience, cognitive science, evolutionary biology and robotics.
Hence the objective of this special session is to assemble a set of
high-quality original contributions that reflect and advance the
state-of-the-art in the area of Evolutionary Robotics, with an emphasis on the
cross-fertilization between ER and the aforementioned research areas, ranging
from theoretical analysis to real-life applications.
Topics of interest include (but are not restricted to):
Evolution of robots which display minimal cognitive behaviour, learning, memory, spatial cognition, adaptation or homeostasis.
Evolution of neural controllers for robots, aimed at giving an insight to neuroscientists or advancing control structures.
Evolution of communication, cooperation and competition, using robots as a research platform.
Co-evolution and the evolution of collective behaviour.
Evolution of morphology in close interaction with the environment, giving rise to self-reconfigurable, self-designing, self-healing and self-reproducing robots or humanoid and walking robots.
Evolution of robot systems aimed at real-world applications as in aerial robotics, space exploration, industry, search and rescue, robot companions, entertainment and games.
Evolution of controllers on board real robots or the real-time evolution of robot hardware.
Novel or improved algorithms for the evolution or robot systems.
The use of evolution for the artistic exploration of robot design.
Exploiting the Computational Properties of the Immune System: Applications and Algorithms
Recent developments in immunology propose a re-positioning of the natural
immune system away from the traditional view of it as purely a defence system
to a complex, self-organising computational system which is able to
continuously compute the state of the body, and then to respond appropriately.
This achieves host regulation, maintenance and of course, protection. This
interpretation of the human immune system emphasises a number of its functional
properties which have clear potential for exploitation in computational and
engineered systems. Those properties encompass embodiment, composition of
heterogeneous and naturally distributed components, life-long and continuous
learning, adaptation and homeostasis. The inclusion of such properties
differentiates the immune-inspired paradigm of Artificial Immune Systems (AIS)
from other biologically-inspired paradigms such as Evolutionary or Swarm
Computation.
Algorithms inspired by the human immune system have been successfully
applied to a variety of application domains, most notably pattern recognition
and optimisation. However, it appears that rich research potential remains in
the exploitation of the unique properties of the immune system for transfer
into the computational domain, which are also essential properties for many
engineered systems. For example, recent advances in technology enable the
construction of pervasive, autonomous systems constructed from perhaps tens of
thousands of devices. Such systems must exhibit autonomic properties including
self-repair and self-optimisation in addition to achieving desired
functionality. For instance, networked devices such as internet routers which
operate in dynamic and unpredictable environments must both maintain integrity
and exhibit fault-tolerance.
This workshop addresses the development and application of novel immune-inspired
algorithms to applications which can benefit from the unique and defining features of
the immune system. In order to progress research in this area, an interdisciplinary
approach to algorithm design is required, in which algorithms are developed that are
rooted firmly in the underlying immunology. This process necessitates mathematical
and computational modelling, in addition to obtaining a strong understanding of the
respective application areas. To this end, we solicit both position and technical papers
which address these areas.
Hardware Aspects of Bio-Inspired Architectures and Systems
Session Organisers: Andrew J. Greensted & Martin A. Trefzer
Bio-inspired techniques and systems, supported by a wide range of state of
the art electronic systems, have the potential for creating novel and
competitive real-world applications. Furthermore, this research area offers the
possibility to explore and master emerging technologies. However, could the
challenge of embedding complex algorithms in hardware or developing
proof-of-principle hardware prototypes into complete solutions explain why they
have not been more widely adopted?
With the rapid increase in computational power of standard low-cost PCs, it
has become easier to develop highly sophisticated algorithms within the field of
evolutionary computation. Due to this, most applications are only developed in
software relying on large systems utilising great computational power. However,
without undertaking a hardware implementation stage, these systems reduce their
chances of being deployed in many real-world applications and loose out on the
advantages customised hardware platforms can offer: reduced size, energy
efficiency, dependability and mobility as well as greater parallelism and
hardware acceleration of critical operations. In order to make novel
bio-inspired techniques industrially and commercially competitive, these
properties are crucial and generally cannot be achieved in software only
solutions. In areas such as robotics, mobile computing, automotive
industries and real-time data processing these factors become vital.
How do people see Hardware?
Tool or Tissue?
Used or Abused?
Explored or Exploited?
This session is intended to bring together researchers who are implementing
bio-inspired techniques in hardware, who are addressing the challenges this
process presents and who are pushing forward alternative technologies for
bio-inspired investigations. This session will provide a great opportunity for
researchers to discuss their approaches and exchange their expertise and
solutions. Submitted papers should be based upon, but not restricted to, the
following topics:
Enhancing Electronic Systems and Advancing Technology with Bio-Inspired Techniques
Businesses require accurate forecasts of demand in order to make effective
decisions, such as marketing, financial investment, inventory, distribution,
human resource planning, purchasing, and so on. These forecasts are usually
based on a function combination system (forecasting with evolutionary computing
models) of traditional statistical methods, evolutionary algorithms (EA),
evolutionary computation (EC), and management judgment. Although the wide
application of hybrid modeling concepts, due to lack of abilities to catch the
forecast data pattern, hybrid evolutionary algorithms resulted in over-reliance
on the use of informal judgment and higher expense.
With the advantages of EA computing capabilities over the traditional
optimization approaches, recently, they have been applied to catch the data
pattern more accurate via systematical computation process, however, hybrid
evolutionary algorithms (HEA), such as genetic algorithms with simulated
annealing algorithms (GA-SA), chaotic search with particle swarm optimization
algorithm (CPSO) and chaotic search with genetic algorithms (CGA), require more
detail researches and empirical studies.
The objective of this special session is to invite together research and
application of hybrid evolutionary algorithms for any forecasting
applications.
This special session invites contributions in all aspects of applying hybrid
evolutionary algorithms to improve the usage efficiency of those algorithms and
aims to promote the discussion and exploration of new ideas. Topics of
interests include (but not limited to):
The usage of HEA in forecasting applications.
Theoretical comparison of HEA and EA in forecasting applications.
Empirical comparison of of HEA and EA in forecasting applications.
Parameter determination by genetic algorithms with simulated
annealing algorithm (GA-SA) in forecasting applications.
Parameter determination by chaotic search with particle swarm
optimization algorithm (CPSO) in forecasting applications.
Parameter determination by chaotic search with genetic algorithms (CGA) in forecasting applications.
Other application of novel HEAs in forecasting applications.
Genetic and evolutionary algorithms (GEAs) are powerful search methods based
on the paradigm of evolution and widely applied to solve problems in many
disciplines. In order to improve the performance and applicability, numerous
sophisticated mechanisms have been introduced and integrated into GEAs in the
past decades. One major category of these enhancing mechanisms is the concept
of linkage, which models the relation between the decision variables with the
genetic linkage observed in biological systems, and linkage learning
techniques. Linkage learning connects the computational optimization
methodologies and the natural evolution mechanisms. Not only can learning and
adapting natural mechanisms enable us to design better computational
methodologies; the insight gained by observing and analyzing the algorithmic
behavior permits us to further understand biological systems, based on which
GEAs are developed.
This special session aims at providing a forum for reviewing of current
state-of-art linkage learning techniques, exchanging of ideas and viewpoints on
linkage, as well as discussing the future directions. We invite researchers to
submit their original and unpublished work including but not limited to the
following topics:
Linkage in biological systems and computational algorithms
Linkage for discrete/continuous variables
Linkage processing, handling, and learning techniques
Identification and utilization of linkage
Adaptation of representation and/or operators for linkage
Theoretical aspects of linkage
Applications of the linkage concept
Position papers
Real-world applications
Memetic Algorithms for Hard to Solve Problems
Session Organisers: Ferrante Neri, Pablo Moscato & Hisao Ishibuchi
One of the recent growing areas in Evolutionary Algorithm (EAs) research is
Memetic Algorithms (MAs). MAs are population-based meta-heuristic search
methods inspired by Darwinian principles of natural evolution and Dawkins
notion of a meme defined as a unit of cultural evolution that is capable of
local refinements. Recent studies on MAs have revealed their successes on a
wide variety of real world problems. Particularly, they not only converge to
high quality solutions, but also search more efficiently than their
conventional counterparts. In diverse contexts, MAs are also commonly known as
hybrid EAs, Baldwinian EAs, Lamarkian EAs, cultural algorithms and genetic
local search.
The aim of this special session is to reflect the most recent advances in
the field, and propose novel algorithmic implementations of MAs oriented
towards specific problem which are hard to solve by classical optimization
methods and popular meta-heuristics. A high emphasis will be given to the
problems of balancing global and local search and on the techniques for
obtaining an efficient coordination of the local search within an evolutionary
framework. Both theoretical and empirical works are in the scope of this
session. Some examples of the aforementioned hard to solve problems by means of
Memetic Computing are:
Dynamic optimization problems
Noisy fitness landscapes
Computationally expensive optimization problems
Large scale problems
Multi-objective problems
Real-world applications
Parallel Bio-inspired Optimisation Methods: Algorithms and Applications
Session Organisers: Dr Andrew Lewis, Dr Sanaz Mostaghim & Dr Marcus Randall
The use of computational models is becoming routine across a wide range of
industries and applications. In the engineering design process, and scientific
research, they are often used to find the best of a number of solutions as
measured by some objective function(s). There is a growing demand for tools
able to perform automatic optimisation to allow rigorous and systematic
exploration of the model, particularly with high-dimensional parameter spaces.
This special session invites papers discussing recent advances in the
development and application of bio-inspired optimisation algorithms to the
field of computational science, optimisation, parallel and Grid computing. We
encourage submission of papers describing new concepts and strategies, and
systems and tools providing practical implementations, including hardware and
software aspects. Papers describing methods of exploiting parallel
computational resources on these algorithms are particularly encouraged.
In addition, we are interested in application papers discussing the power and
applicability of these parallel methods to real-world problems in both
well-established areas, such as computational engineering, and emerging fields
such as computational biology.
Evolutionary computing researchers and practitioners know very well that
choosing good parameter settings is essential for good performance of
evolutionary algorithms. However, the defining attributes (e.g., the parent
selection method) and the parameter values (e.g., the mutation rate) of
evolutionary algorithms have been and often still are chosen in an ad hoc
fashion, regularly on the basis of unverified conventions and beliefs.
Adaptation of parameters and operators through tuning (off-line) or control
(on-line) comprises a field of research that seeks to address this omission. As
such, it is currently one of the most important and promising areas of research
in evolutionary computation. The aim of this special session is to reflect the
state-of-the-art in the field, and raise the awareness of this important area
in the evolutionary computation community.
Topics of interest include, but are not limited to:
Parameter Tuning
Parameter Control
Adaptive Parameter Control
Deterministic Parameter Control
Self-Adaptation
Meta Algorithms
Adaption of Representation
Adaption of Variation Operators
Adaption of Selection Operators
Adaption of Fitness Function
PerAda: Adaptation Strategies for Pervasive Adaptation
Session Organisers: Ben Paechter, Emma Hart & A.E. Eiben
The field of Pervasive Adaptation - PerAda - is concerned with researching novel design
paradigms for massive-scale pervasive information and communication systems which will
enable a technology rich-future in which computing is truly ubiquitous. Such systems will
operate in an ever-changing networked environment and will have to continuously and
autonomously organise and adapt to highly dynamic and open technological and user
contexts.
This workshop addresses the use of adaptation strategies in pervasive systems. Adaptation
strategies, which may be bio-inspired, stochastic or otherwise, will need to operate over
different time scales and speeds, ranging from short term adaptation to long-term evolution.
This impacts the entire spectrum of research in Pervasive Systems in that it will imply
changes in software, hardware, protocols and/or architecture at different levels of granularity
and abstraction. Adaptation must occur at the level of individual devices as well as in 'tribes'
of artefacts which are formed on an ad-hoc basis; this is compounded by that fact that the
composition and location of systems is dynamic and continuously subject to change.
Evolution and adaptation in such environments poses a number of challenges. Evolution
must occur within the boundaries imposed by ensuring trust and security in the networks, and
further more, the potential for 'runaway evolution' must be addressed: any decentralised self-
organising system which enables information or ideas to be propagated is vulnerable to being
overcome by memes that are prevalent because of their ability or tendency to reproduce
rather than because they are useful. A meme might be a piece of information, some code, a
grouping or structure of entities etc. The high prevalence of some less useful applications and
groupings in Facebook is an example of this. This behaviour in the system may result in an
undesirable signal/noise ratio or ultimately to bandwidth saturation. New decentralised
mechanisms need to be developed to prevent, monitor, evaluate and control these memes
with a viral (but not necessarily malicious) nature.
Papers are welcomed which includes any aspect of adaptation in a Pervasive Environment.
We welcome both technical and position papers. Examples of topics include, but are not
limited to:
Evolution/Adaptation of hardware in pervasive systems
Evolution/Adaptation of software
Evolution of societies of artefacts
Runway evolution
Adaptation on multiple time-scales
Novel paradigms for continuous adaptation of devices
Exploitation of memory mechanisms for efficient adaptation
Human-intervention in adaptation and evolution of devices
Enabling self-* properties in pervasive systems.
Performance Assessment of Constrained / Bound Constrained Multi-Objective Optimization Algorithms
Session Organisers: Qingfu Zhang & P. N. Suganthan
Optimization for multiple conflicting objectives results in more than one optimal solutions known as Pareto-
optimal solutions. Although one of these solutions is to be chosen eventually, the recent trend in evolutionary
multi-objective optimization studies have focused on approximating the Pareto front by a set of solutions.
Such a set of solutions can collectively provide a good insight to the different trade-off regions on the
resulting efficient frontier, thereby aiding a better and more confident decision making.
Evolutionary multi-objective optimization (EMO) methodologies have been suggested since the eighties for
this task. Since then a number of performance assessment methods has also been suggested. After more
than 20 years of research and development of efficient EMO algorithms, we realize that it is time to evaluate
the existing EMO methodologies on carefully chosen test problems which are scalable with respect to the
objectives, the decision variables and constraints with complex Pareto shape in the decision space. The
comparisons will be made for a limited number of overall evaluations, so that the existing or new algorithms
can be evaluated for different functional abilities:
convergence to Pareto front with diversity,
to scale well on many objectives,
to scale well on many variables,
to perform well with bound and general constraints
able to tackle complex Pareto shape in the parameter space
able to tackle varying degree of linkages among variables.
PERPLEXUS: Pervasive reconfigurable platform for modelling complex systems
Session Organisers: Andres Upegui & Andres Perez-Uribe
The simulation of large-scale complex systems requires huge amounts of computing
resources. Even if the Moore law yet guarantees increasingly more powerful chips every year,
the capacity of a single device is not always able to provide the optimal solution for running
this kind of application. A new kind of distributed computing could appear in the near future,
with the advent of a new era of computing, after the mainframe era, and the PC era. In this 3rd
era of computing, we are likely to see a myriad of ubiquitous devices with large computing
and sensory capacities in our environment.
Anticipating this new era of computing, we can imagine a solution for simulating complex
systems, using the computational resources of ubiquitous devices in our environment. These
devices will be adaptable and retargetable, for this, they will be made of custom
reconfigurable devices endowed with adaptation capabilities that will enable the simulation of
large-scale complex systems and the study of emergent complex behaviors in virtually
unbounded wireless networks of computing modules.
The aim of this special session is to bring together hardware, middleware, and application
developers of systems that exploit the pervasiveness of ubiquitous systems in order to model
complex systems. This session will allow researchers to share experiences and identify
theoretical and technical issues. Submitted papers may describe applications, computing
models, modeling frameworks, or hardware platforms.
Techniques for Online and Distributed Evolutionary Computation
Session Organisers: Dr Daniele Miorandi, Dr Lidia Yamamoto, Dr Emma Hart & Dr Tina Yu
Within the context of computing and communication the 'complexity ceiling'
limits the ability to introduce innovation and to cope with changing user needs
and demand. In order to overcome such limitations, it is desirable for many
software and hardware systems to embed adaptation and evolution capabilities in
the system fabric itself. By doing so, they would be able to work and perform
well under an extreme variety of operating conditions, while at the same time
easing system management tasks. Online evolution is difficult to achieve due
to a variety of problems and challenges. These include the need to envisage
extremely resilient evolutionary mechanisms (able to evolve without disrupting
the system operations), the ability to devise new strategies (in response to
external stimuli), and to operate in noisy environments. In other words, an
effective on-line system needs to continuously provide evolvability to cope
with an open- ended changing environment.
Extra challenges are also faced due to the interconnected and distributed
nature of many systems. Such systems cope with only partial information (as
single nodes/clusters may not be aware of the global system status), and in
many cases with delayed information on the (estimated) fitness level of the
current solution. In addition, distributed systems are often composed of
heterogeneous devices, perhaps operating over different timescales and with
different constraints, further increasing the challenges of achieving evolution
in the global system.
Topics of interest include, but are not limited to:
resilient online and distributed EC mechanisms
robustness vs. evolvability in online environments
online evolution in artificial chemistries and chemical computing
online/distributed evolutionary optimization in uncertain environments
decentralized online evolutionary techniques
online/distributed evolution in robotics and embryonic circuits
online evolution in networked, distributed and pervasive systems
Theoretical Foundations of Evolutionary Computation
Session Organisers: Benjamin Doerr & Frank Neumann
Evolutionary computation methods such as evolutionary algorithms or ant colony
optimization have been shown to be very successful when dealing with real-world
applications or problems from combinatorial optimization. The theoretical under-
standing of these, in practice successful, algorithms is an important topic and has
gained increasing interest in recent years. The aim of this special session is to bring
together people working on theoretical aspects of evolutionary computation meth-
ods. We invite submission concerning all kinds of theoretical analyses of evolutionary
computation. Topics of interest include (but are not limited to):
population dynamics
approaches from statistical mechanics
runtime analysis
fitness landscapes and problem difficulty
self-adaptation
What is Computational Swarm Intelligence?
Session Organisers: T. Blackwell, M. Bishop & Slawomir Nasuto
Swarm Intelligence (SI) algorithms, often inspired by communication and
interaction between social agents such as ants or bees share much in common
with EA's. However the precise relationship remains vague and ill-defined.
Are there any principles underlying and distinguishing the behaviour of swarm algorithms?
The session will gather experts in adjacent and seemingly related fields of
Swarm Intelligence
Artificial Immune Systems
Multi-agent systems
in order to discuss issues such as overlaps between algorithms, their compatibility, and their essential differences as problem solvers.
Previous research directions have looked towards Physics and, in particular,
Biology for new ideas. This session proposes that Computational Swarm
intelligence is an autonomous aggregate of techniques that so far have not been
unified. We are looking for a mathematical, algorithmic framework which will
enable us to understand and analyse these algorithms.
The aim of this workshop is NOT to make comparisons between techniques, but
rather to make comparisons at a conceptual level.
The session will seek to define the metaheuristics of Swarm Intelligence
algorithms. A common framework is desirable for a number of reasons, including
the following:
Better understanding of the limits of application of SI techniques.
Hope to further the analysis of all algorithms by finding overarching principles.
Suggestions for novel and hybrid algorithms may result from the clash and the synthesis of ideas.
The workshop aims at addressing such issues from an explicitly theoretical
rather than a heuristic perspective. The following contributions are
welcomed:
Position papers and reports of work in progress
Papers proposing and advancing the metaheuristics of popular SI techniques such as PSO, ACO and SDS
Contributions from adjacent fields e.g AIS, Multi-Agent Systems
