Fullscreen
Loading...
 

Print



We invite you to submit a paper to the Generative and Developmental Systems (GDS) track at GECCO 2013, the premier conference for GDS-related work worldwide. GDS covers all aspects of evolved complex systems, with work ranging from biologically inspired approaches to automated engineering design. Each paper submitted to the GDS Track will be reviewed by experts in the field. The size and prestige of the GECCO conference will allow many researchers to learn about your work, both at the conference and via the proceedings (GECCO has the highest impact rating of all conferences in the field of Evolutionary Computation and Artificial Life).

— R. Doursat, M. E. Palmer and J. Bongard, 2013 GDS track chairs

Overview

Complex Systems

Engineering is torn between an attitude of strong design and dreams of autonomous devices. It wants full mastery of its artifacts, but also wishes these artifacts were much more adaptive or "intelligent". Still today, our most sophisticated computer and robotic systems must be spoon-fed at every stage: programmed, repaired, upgraded. Meanwhile, insatiable demand for functional innovation has created an escalation in system size and complexity. In this context, the tradition of rigid top-down planning and control in every detail is becoming unsustainable.

Complex systems—large sets of elements interacting locally and giving rise to collective behavior, can be a powerful source of inspiration toward novel methods and technologies. Understanding such natural systems by modeling and simulation can help create a new generation of truly autonomous and adaptive artificial systems, which would display "self x" properties mostly absent from traditional engineering. Historically, along these lines, the observation of neurons and genes has inspired neural networks and genetic algorithms. On the other hand, these disciplines have also largely shifted their focus to classical optimization and search problems, away from distributed, emergent computation.

More is More

Yet, one of the remarkable properties of evolution is that it can act at any scale. Combining natural selection with self-organization, evolution has produced the astounding complexity and diversity found in living organisms. Thus, in principle, in silico evolution has the potential to reproduce a similar feat and create artifacts of great complexity and diversity—so why hasn't this happened yet?

The GDS track is dedicated to unlocking the full potential of evolutionary computation as a design methodology that can scale to systems of great complexity. It aims to create complex and diverse artifacts that meet our specifications with minimal guidance and programming effort. For this, it promotes the following ideas:

Representations should be indirect and open-ended In the vast genotype spaces of complex systems, some lineages of organisms are better than others at fostering useful innovation. In this sense, a genotype is "successful" not only for its short-term fitness but also its long-term evolvability through its influence on the production of variation. More than the information needed to produce a single individual, it is a layered, modular and reusable repository of many generations. Artificially evolved complex systems, too, must acquire open-ended genotypes with superior evolvability. "Indirect representations" such as morphogenesis or string-rewriting grammars, which rely on developmental or generative processes, create a mapping from genotype to phenotype. This may allow a gradual improvement of long-term evolvability via accumulated elaborations and emergent, unplanned features. In contrast, "direct representations" are not capable of open-ended elaboration because they restrict genotype space to predefined features.

Complex environments encourage complex phenotypes As expected, complex genotypes are not necessarily favored in simple environments. However, complex genotypes also enable original, unprecedented phenotypes and behaviors that cannot be reached by simple genotypes. Therefore, if a more complex phenotype later successfully invades a new, uncrowded niche in a complex environment, it can build pressure toward increasing complexity. Many factors may affect environmental complexity, hence genotypic complexity: they include spatial structure, temporal fluctuations, and competitive co-evolution.

More generations, individuals, and internal components are required Another factor missing from artificial evolution may be that today's typical numbers of generations, sizes of populations, and variables inside inviduals are still too small. Just like physics needs higher-energy accelerators and farther-reaching telescopes to understand matter and space-time, evolutionary computation could use a strong boost in computational resolution and scope to understand the spontaneous generation of complex functionality. It may be no coincidence that the evolution of biological diversity involved four billion years and untold numbers of organisms, some made of trillions of cells.

Evo-Devo Engineering

Over 150 years after Darwin's and Mendel's work, and the subsequent "Modern Synthesis" of evolution and genetics, the developmental transformation from genotype to phenotype is still unclear. Despite the unraveling of the nature of DNA and the molecular basis of gene expression, most biological evolutionary research has focused on selection, and it is only recently that a deeper understanding of the mechanisms of variation became an inportant concern. Evolutionary development, or "evo-devo", is a recent field of biology comparing the morphogenesis of different species at both genetic and anatomical levels. For evo-devo, development cannot remain an abstraction if we want to break the glass ceiling of evolutionary novelty. From what fine-grain material, i.e. combinations of elementary "blocks", do new and functional complex structures spontaneously arise?

Likewise, looking at the full evo-devo picture should be a primary concern of systems engineering and computer science when venturing in the arena of autonomous and adaptive architectures. The ambition of the GDS track at GECCO 2013 is to contribute to new avenues in evolutionary computation by stressing the importance of fundamental laws of generative and developmental variations that generate the raw material for the selection of artificial systems. Because indirect mappings promote compact encodings, modularity and combinatorial reuse, complex multi-agent architectures are uniquely capable of producing ongoing open-ended innovation.
By putting a special emphasis on the evolution of complex systems, the GDS track reaches out to emergent, innovative communities that share the same spirit as artificial embryogeny and rewriting systems, such as: amorphous computing, autonomic networks, collective robotics, evo-devo modeling, morphogenetic engineering, organic computing, reconfigurable robots, spatial computing, swarm chemistry, unconventional computing, natural computing, and many others.



The track and conference will be held in Amsterdam, The Netherlands. For more information, please see the GECCO 2013 homepage at http://www.sigevo.org/gecco-2013. GECCO is sponsored by the Association for Computing Machinery (ACM) Special Interest Group on Genetic and Evolutionary Computation (SIGEVO). SIG Services: 2 Penn Plaza, Suite 701, New York, NY, 10121, USA, 1-800-342-6626 (USA and Canada) or +212-626-0500 (Global).




Contributors to this page: Rene Doursat and webmaster .
Page last modified on Tuesday 14 May, 2013 05:48:13 by Rene Doursat.