Search

The evolutionary nature of humans requires agent systems to be continuously replaced due to their inability to meet or adapt to our changing needs. Therefore, to eliminate the need for a human to continuously adapt an agent, evolutionary agents are required [Chu04, Ore99, Rak02, Syc96]. This dissertation develops a feasible option to ensuring that agents continuously develop desirable behaviour. The solution is a specialized architecture that embeds self-evolvement into a target agent. The specialized architecture ensures that desirable behaviour emerges from any agent, as it is embedded between the target agent and the target agent’s environment and therefore is able to obtain domain- and hardwarespecific information from the target agent. The specialized architecture is a comprehensive methodology that incorporates all agents with the ability to embed the required self-evolvement enhancements as domain- and hardwarespecific information is obtained from the target agent. The specialized architecture responsible for embedding self-evolvement into an agent is the generic self-evolvement effecting evolutionary agent (GSEEA). The GSEEA is developed with a single goal, which is to ensure that the target agent meets the requirements of a changing environment. Changing environmental conditions can include different network conditions and different platforms. The GSEEA’s goal is accomplished by embedding the required self-evolvement enhancements into the target agent to produce a self-evolvement enhanced agent. In this dissertation the GSEEA is implemented to demonstrate its feasibility and problem-solving accuracy. In the GSEEA implementation the target agent is a puzzle-solving agent and the self-evolvement enhanced agent is the selfevolvement enhanced puzzle-solving agent. The GSEEA’s deliberative component consists of two algorithms, namely a genetic algorithm and a learning algorithm. The GSEEA’s genetic algorithm develops knowledge base rules (selfContents III evolvement enhancements) that modify actuator information. The GSEEA’s learning algorithm updates developed knowledge base rules by modifying sensor information. The GSEEA tests the developed self-evolvement enhancements by embedding them into the target agent through the target agent’s knowledge base manager, evaluating the developed self-evolvement enhancements and deleting those which do not enhance the target agent. The target agent achieves selfevolvement as additional enhancements required by the self-evolvement enhanced agent can be achieved by applying the same process followed to enhance the target agent which was discussed previously. The evaluation of the GSEEA implementation demonstrated that the GSEEA was implemented successfully based on feasibility and problem-solving accuracy as the self-evolvement enhanced puzzle-solver agent outperformed the puzzlesolver agent. Prof. E.M. Ehlers

Tez (Yüksek Lisans) -- İstanbul Teknik Üniversitesi, Bilişim Enstitüsü, 2006, Thesis (M.Sc.) -- İstanbul Technical University, Institute of Informatics, 2006, Gösterilimlerin evrimsel algoritmalar üzerindeki etkisi durağan ortamlar için bugüne kadar birçok çalışmada incelenmiş, bununla beraber dinamik ortamlardaki etkisi ihmal edilmiştir. Bu çalışmada, farklı gösterilimlerin dinamik ortamlardaki etkisini deneysel olarak inceledik. Probleme ara dönüşüm olmaksızın çözüm olabilen gösterilimlere, dogrudan gösterilimler denir. Dolaylı gösterilimlerde ise, arama farklı bir uzayda yapıldığından dolayı, çözümün doğrudan gösterilim haline ulaşmak için bir dönüşüm gerekir. Genelde, doğrudan gösterilimler tüm uzayda arama yaptıklarından dolayı geçersiz çözümlerle de evrim sırasında baş etmelidir. Dolaylı gösterilimler genetik arama uzayında arama yaptığı ve aday çözümler genelde geçerli çözüm uzayına izdüştüğü için, geçersiz çözüm problemi yoktur. Testlerde çoğunluk tarafından bilinen çok boyutlu sırt çantası (multi- dimensional knapsack problem) ve gezgin satıcı (traveling salesman) optimizasyon problemleri için doğrudan ve dolaylı yöntemleri inceledik, ve karşılaştırdık. Çok boyutlu sırt çantası problemi için seçtiğimiz dolaylı ağırlık kodlama (weight coding) yöntemi ile travelling salesman problemi için kullandığımız dolaylı ötelenmis koordinatlar (perturbed coordinates) yöntemi, çözümü bulmak için temelde aynı mantığı paylaşıyor. Her iki yöntemde de aday çözümler, orjinal problemin bazı değerlerini değiştirerek biraz farklı bir problem elde ediyor. Daha sonra elde edilen yeni probleme hızlı bir sezgisel yöntemle çözüm buluyor. Bulunan bu çözümü de orjinal problemin çözümüymüş gibi kullanıyor.Dolaylı gösterilimlerde, her değişim anında, toplumu oluşturan çözümlerin genetik arama uzayındaki bileşenleri, çözüm uzayına izdüşürülerek, toplumun yeni probleme adapte olması sağlanıyor. Sonuçlar dolaylı gösterilimlerin değişim anlarında, sahip oldukları sezgisel adaptasyon mekanizmasıyla var olan çözümleri yeni probleme adapte etmelerinden dolayı dinamik problemler için daha uygun olduğunu gösterdi. Ek olarak, gösterilimlerin dinamik ortamlardaki etkisinin, statik ortamlardakinden daha büyük olduğunu gördük. Bu nedenle dinamik ortamlarda gösterilimler seçilirken seçici olunmalı, adaptif yöntemler tercih edilmelidir., The effect of different representations has been thoroughly analyzed for evolutionary algorithms in stationary environments. However, the role of representations in dynamic environments has been largely neglected so far. In this study, we empirically analyze the effects of different representations in dynamic environments. A representation is called direct, if it can be interpreted directly as a solution to the problem. In indirect representations, search is done in a different space, so mapping is required to get the direct representation of the solution. Direct representations do search in the entire search space, so generally method should deal with the infeasible solutions through evolution. In indirect representations, search is done in genetic search space and generally they all map to feasible solutions of the search space. Thus indirect representation do not have a problem of infeasible solutions. In tests, we analyzed and compared, direct and indirect representations for both generally known optimization problems, multi-dimensional knapsack problem and traveling salesman problem. Two indirect representations selected, weight coding for multi-dimensional knapsack problem and perturbed coordinates for traveling salesman problem, keep the same track while finding the solution. In both approaches, candidate solutions first creates a slightly changed problem, by changing some values of the original problem. Then they find a solution to the changed problem using a fast heuristic. In indirect representations, at each change point, existing population is adapted to the new problem by mapping solutions' components of genetic search space to solution space. Our results indicate that indirect representations are particularly suitable for dynamic problems, because they implicitly provide a heuristic adaptation mechanism that improves the current solutions after a change. In addition, we saw that the choice of representation in dynamic environments is even more important than in static environments. For this reason, you should be selective while deciding representation to be used in a dynamic environment and you should prefer adaptive representations., Yüksek Lisans, M.Sc.

A new evolutionary design tool is presented, which uses shape grammars and a grammar-based form of evolutionary computation, grammatical evolution (GE). Shape grammars allow the user to specify possible forms, and GE allows forms to be iteratively selected, recombined and mutated: this is shown to be a powerful combination of techniques. The potential of GE and shape grammars for evolutionary design is examined by attempting to design a single-person shelter to be evaluated by collaborators from the University College Dublin School of Architecture, Landscape, and Engineering. The team was able to successfully generate conceptual shelter designs based on scrutiny from the collaborators. A number of avenues for future work are highlighted arising from the case study. Science Foundation Ireland Irish Research Council for Science, Engineering and Technology 6 months embargo - could not locate month of publication: publish in December to be safe? - AV 29/10/2010 ti, ke, published in April 24th according to http://www.ingentaconnect.com/content/ind/ijde/2010/00000003/00000001/art00001;jsessionid=hlwhjx5unv57.alexandra - AS 04/11/2010

RESUMEN: En el campo de la Inteligencia Artificial se han propuesto técnicas en tres tipos básicos de aprendizaje: el aprendizaje supervisado, el aprendizaje no supervisado y el aprendizaje por refuerzo. Las implementaciones que se hacen de estas técnicas están enfocadas en su mayoría a un aprendizaje computacional que poca interacción tiene con el humano; es un aprendizaje que se compara con experiencias anteriores pero no se logra a partir de la propia interacción máquina-humano. Evol-PlayMac surge, entonces, como un sistema que trata en lo posible de solucionar este problema al tratarse de un desarrollo que tiene implementada una técnica de aprendizaje del área de la computación evolutiva, que permite que a través de un juego tradicional de tablero, el parqués, la máquina formule estrategias de juego que cada vez demuestren un mejor desempeño obteniendo un nivel de ganancia mayor y logrando esto a partir de la competencia contra un humano.