GECCO-99
Bird-of-a-feather Workshop Summaries
GECCO-99
Bird-of-a-feather Workshop Summaries
For information regarding participating or presenting at a particular workshop, please see the workshop homepage. For general inquiries regarding workshops, please contact Annie Wu at aswu@aic.nrl.navy.mil. The workshop schedule is available here. Information for workshop organizers is available here.
Note about hotel availability on July 12, 1999.
The joint
AAAI-99
and GECCO-99 Workshop will be held on July 18, 1999
and is open to both GECCO-99 and AAAI-99 attendees.
Because there is limited attendance at this joint workshop,
please check with the organizer for both participation and attendance details.
Many real-world optimization problems are eventually dynamic. New jobs are to be added to the schedule, the quality of the raw material may be changing, new orders have to be included into the vehicle routing problem etc. In such cases, when the problem changes over the course of the optimization, the purpose of the optimization algorithm changes from finding an optimal solution to being able to continuously track the movement of the optimum through time.
Since natural evolution is a process of continuous adaptation, it seems straightforward to consider evolutionary algorithms as appropriate candidates for dynamic optimization problems. And indeed, the interest in evolutionary algorithms for dynamic optimization problems is growing and a number of authors have proposed an even greater number of new approaches. However the field seems to lack a general understanding as to suitable benchmark problems, fair comparisons, or measurement of algorithm quality.
The goal of the workshop will be to foster interest in the subject, get together researchers working on that topic, and achieve an informal agreement on some of the key issues in the field.
Evolutionary Computation researchers have long had an interest in parallel processing, due to the ease with which many EC algorithms can be implemented in a parallel fashion. Furthermore, many real-world applications of EC require some parallel processing in order to make practical progress. This workshop invites discussion on general issues of parallel processing in the context of EC, including (but not necessarily limited to):
Software visualization is an area of computer science devoted to supporting people's understanding and effective use of computer software. Within Evolutionary Computation, visualization is increasingly being used as a means for supporting people's understanding of their algorithm's behavior. This workshop offers a timely opportunity for people interested in the possibilities of visualization to see the current state of the art, discuss the contribution that this work makes, and influence the future work done in this area.
This workshop asks participants to identify and discuss strategic issues in methodology, pedagogy and philosophy that are critical to the future of genetic and evolutionary computation. The particular focus for the workshop this year will be the topic "reporting and research practices." The workshop participants are asked to:
For last five years or so, the research on multi-criterion optimization using evolutionary algorithms (EAs) has received a growing attention. This is rightly so, because population-based EAs allow a unique way to find multiple Pareto-optimal solutions in multi-criterion search and optimization problems. To date, there exists successful EA implementations and a number of engineering applications in this area. In this workshop, we intend to present the research results that have been achieved already and discuss how and what topics should be emphasized for future research:
There has been a growing interest in data mining in several AI-related areas, including evolutionary algorithms. Hence, it seems that it is the right time for the communities of data mining and evolutionary algorithms to meet and exchange ideas. The general goal of the workshop will be to discuss promising and necessary research directions in data mining with evolutionary algorithms.
Topics include:
The basic genetic programming paradigm involves the use of genetic representations and genetic manipulations based upon simple context-free grammars and Koza's closure property. A number of researchers have recently examined variations to these aspects of GP. In particular, some researchers have relaxed the closure property (e.g., Strongly Typed Genetic Programming) while others have considered more complex grammar types (e.g., logic grammars and attribute grammars). As well, in EC in general, a variety of grammar techniques have been used (e.g., Lindenmayer systems).
Is vanilla GP sufficient for the evolution of solutions to all complex problems? Obviously not. It is therefore important for us to consider variations upon GP which may improve its applicability to hard problems. Incorporating increased complexity into the genetic representations and manipulations is one obvious course of exploration. The various advanced grammar techniques being used for such exploration face common issues and are drawing similar conclusions. Rather than continuing to tout each technique independently (e.g., STGP, cellular encoding, etc), the EC community should try to discover what these techniques have in common and where they are leading us. This workshop aims to present a cross-section of current research and promote a discussion of fundamental issues concerning grammar design in evolutionary algorithms.
In solving complex tasks using evolutionary methods there are often many conflicting requirements, and there are many complexities in breaking down a complex problem into effective subproblems. It has been proposed that such complexities can be handled by coevolutionary interactions, such as cooperation, competition and host-parasite or predator-prey interaction.
This workshop will consider ways in which these coevolutionary processes can be applied to improving the performance of evolutionary algorithms, both in improving the performance of algorithms in existing problem domains and using the techniques to solve problems which are not amenable to conventional evolutionary computing techniques. Combining evolutionary computing ideas with the problem decomposition and distributed solution techniques which lie at the heart of innovations such as agent technology provide a powerful paradigm for problem solving in many domains. These techniques provide a cyber-diversity of problem-solving methods which echo the bio-diversity found in nature.
The goal of this workshop is to discuss the recent developments in learning classifier systems (LCSs) research and the expected trends of the field. LCSs were introduced by John Holland (1978) as a method of learning by interacting with an environment, based on a biological metaphor: learning is viewed as a process of ongoing adaptation of an agent to an initially unknown environment. For a long time the LCS paradigm has been considered limited to the evolutionary computation community and their applications were limited to well-defined fields, e.g. robotics. The recent developments in the field of reinforcement learning have brought new attention to the LCS paradigm, which has been shown to be an interesting alternative to traditional reinforcement learning techniques. Furthermore, LCSs can be competitive in more general contexts like: autonomous agents, classification, trading agents, and personal assistants. Because of these new developments it is important to bring together people from this field for getting an overview of the latest results and most promising research directions.
"Mathematical and computational approaches to biological questions, a marginal activity a short time ago, are now recognized as providing some of the most powerful tools in learning about nature; such approaches guide empirical work and provide a framework for synthesis and analysis." (Levin et al. 1997) This workshop will be devoted primarily to configuration/individual-based/agent-based models and their connections to both analytical theory and experiment. The workshop will be focused a combination of results along with methodological issues and approaches to the problem of contributing to biology through theory. Methodological issues include: Determining what to put in and what to leave out of a model, representation and implementation of biological systems, testing the accuracy of a model, and applying the results to the real world. We anticipate contributions in, but not limited to, evolutionary theory, ecological theory, immunology, paleobiology, and epidemiology.
Levin, S. A., B. Grenfell, A. Hastings, and A. S. Perelson. 1997. Mathematical and computational challenges in population biology and ecosystems science. Science 275:334-343.
Evolvability is a central issue in evolutionary computation, but remains little understood. It has attracted some interest among workers in evolutionary computation (examples are Altenberg 1994; Wagner & Altenberg 1996) and biology (Dawkins 1989; Kirschner & Gerhart 1998). Much remains to be done, both in understanding the nature of evolvability and working out how to enhance evolvability for the benefit of evolutionary algorithm performance. This workshop will focus on understanding evolvability, how to quantify its characteristics, and how to exploit evolvability to improve algorithm performance.
Altenberg, L. 1994 The evolution of evolvability in genetic programming. In: Advances in Genetic Programming, K. E. Kinnear Jr., (ed.). MIT Press.
Dawkins, R. 1989 The evolution of evolvability. In: Artificial Life, C. Langton, (ed.). Addison-Wesley.
Kirschner, M. & Gerhart, J. 1998 Evolvability. Proceedings of the National Academy of Sciences of the USA 95: 8420-8427.
Wagner, G. P. & Altenberg, L. 1996 Complex adaptations and the evolution of evolvability. Evolution 50: 967-976.
Producing useful theoretical results in genetic programming has been very difficult, although the situation is improving now. In the workshop we want:
In natural evolution one finds impressive examples for the principle of exploiting new sensory channels and making use of the implicit information they encode. Olfactory, tactile, auditive and visual, but also e.g. electric and even magnetic senses have emerged in a vast multitude of variants, often utilizing organs not originally "intended" for the purpose they serve at present. Motivated by these observations, the topic of sensor evolution is becoming a very modern and promising direction of research situated between biology, robotics and Artificial Life. Research in this direction strives at:
1. insights into how biological systems evolve strategies to access new information channels
2. new concepts for design of sensors for flexible and adaptive autonomous agents, an important issue in evolutionary robotics
3. an understanding of the relationship between the information available to an agent and the way it is processed, which is of particular interest for Artificial Life research.
These questions can be approached by studying biological systems as well as hardware or software realizations of evolvable sensors.
Telecommunications is a vital and growing area, important not only in its own right, but also for the service it provides to other areas of human endeavour. However, telecommunications, supported by an ever-changing set of technologies and providing an ever-expanding set of services, presents a challenging range of difficult design and optimisation problems. In recent years, there has been increasing interest in the application of evolutionary computation (EC) to these problems, including network design, call routing, signal processing, frequency assignment, wavelength allocation, capacity planning, admission control, network management, and many others.
Recognising the potential of evolutionary telecommunications, as this synergy has been styled, the workshop will focus on EC for telecommunications applications, particularly those involving areas unique to telecommunications, such as network design. The two main aims will be to establish the current state of the art in the field, as well as to discuss useful directions for future research in this important area.
This workshop will attempt to clarify the outstanding unresolved issues in DNA computing. Issues and opinions are solicited. A brief and incomplete list of concerns is given below. The organizer would very much like to to have other issues raised, and suggestions for the format of the workshop.