Habitat loss and fragmentation have been identified as the main drivers of biodiversity loss. These drivers increase the proportion of habitat edges and change the configuration of landscapes. Habitat edges are known to affect ecological patterns and processes, however, is still unknown how these boundaries affect the assemblage of interactions among species within a community, and particularly its structure. Food webs depict not only the composition of the community, but also the feeding links, which represent a measure of energy flow. Therefore, they can inform about the relationships among community diversity, stability, and ecosystem functions. This thesis explores the effects of habitat edges across native vs. managed forests on the food web of a tri-trophic system comprising plants, herbivores (Lepidoptera larvae) and predators (parasitoids). Particularly, it addresses three main objectives: 1) how food webs at habitat edges are assembled from the species and interactions present in the adjoining habitats; 2) how phylogenetic diversity and the coevolutionary signal among interacting species change across a habitat edge gradient; and 3) whether the mechanisms driving community-wide consumption rates and the ecosystem service of pest control are related to structural characteristics of the food webs. The key findings of this thesis are that, despite the composition of species and interactions of native and managed habitats merging at their interface, food-web structure did not arise as a simple combination of its adjacent habitat webs, potentially due to differential responses of organisms to habitat edges. Moreover, beyond taxonomic composition, the phylogenetic diversity and signal of coevolution among interacting species also change between habitat types, even though this did not translate to changes in consumption rates. Consumption rates and their stability increased with complementarity and redundancy in resource-use among predators. This reflects how environmental changes such as habitat fragmentation can have an effect beyond composition per se, affecting the assemblage of species interactions and even potentially interfering with natural evolutionary processes. Therefore, using interaction-network approaches for determining the impacts of changes may shed light on the underlying mechanisms driving such changes, and help to develop landscape management plans that reduce negative effects on species assemblages.

Some populations of newts (Taricha granulosa) possess large quantities of the neurotoxin tetrodotoxin (TTX) in their skin and eggs. Many populations of garter snake (Thamnophis sirtalis) are resistant to this toxin and can consume large numbers of newts with no negative effects. Despite the wealth of information acquired on the interaction between newts and their predator, garter snakes, very little research has been conducted on possible interactions between newts and other predators. I conducted a suite of experiments examining for the presence of other predators on newts, specifically focusing on predators of their eggs and larvae. I found a single predator, caddisfly larvae were capable of consuming the toxic eggs. Larval caddisflies are extremely abundant at one study site (775,000 caddisfly larvae per pond), and appear to be resistant to the negative effects of ingesting tetrodotoxin. After hatching, larval newts retain substantial quantities of TTX and most are unpalatable to predatory dragonfly naiads. Ovipositing female newts respond to the presence of caddisflies by depositing their eggs at the top of the water column where they are out of the reach of most predatory caddisflies. When caddisflies do consume a newt egg, some of the toxin is retained in their body tissues. Finally, caddisflies consume more newt eggs when those eggs contain less toxin versus eggs that contain large amounts of TTX. This may cause newt eggs that contain low quantities of TTX to more likely to die of predation which could ultimately drive an increase in toxicity of the adult population over time. Collectively, these findings indicate an additional player, caddisfly larvae, is a major predator of newts and could be involved in the evolution of tetrodotoxin toxicity in newts.

[pt] Na área de Exploração e Produção (EeP) de petróleo, uma das tarefas mais importantes é o estudo minucioso das características do reservatório para a criação de modelos de simulação que representem adequadamente as suas características. Durante a vida produtiva de um reservatório, o seu modelo de simulação correspondente precisa ser ajustado periodicamente, pois a disponibilidade de um modelo adequado é fundamental para a obtenção de previsões acertadas acerca da produção, e isto impacta diretamente a tomada de decisões gerenciais. O ajuste das propriedades do modelo se traduz em um problema de otimização complexo, onde a quantidade de variáveis envolvidas cresce com o aumento do número de blocos que compõem a malha do modelo de simulação, exigindo muito esforço por parte do especialista. A disponibilidade de uma ferramenta computacional, que possa auxiliar o especialista em parte deste processo, pode ser de grande utilidade tanto para a obtenção de respostas mais rápidas, quanto para a tomada de decisões mais acertadas. Diante disto, este trabalho combina inteligência computacional através de Algoritmo Genético Co-Evolutivo com Geoestatística de Múltiplos Pontos, propondo e implementando uma arquitetura de otimização aplicada ao ajuste de propriedades de modelos de reservatórios. Esta arquitetura diferencia-se das tradicionais abordagens por ser capaz de otimizar, simultaneamente, mais de uma propriedade do modelo de simulação de reservatório. Utilizou-se também, processamento distribuído para explorar o poder computacional paralelo dos algoritmos genéticos. A arquitetura mostrou-se capaz de gerar modelos que ajustam adequadamente as curvas de produção, preservando a consistência e a continuidade geológica do reservatório obtendo, respectivamente, 98 por cento e 97 por cento de redução no erro de ajuste aos dados históricos e de previsão. Para os mapas de porosidade e de permeabilidade, as reduções nos erros foram de 79 por cento e 84 por cento, respectivamente. / [en] In the Exploration and Production (EeP) of oil, one of the most important tasks is the detailed study of the characteristics of the reservoir for the creation of simulation models that adequately represent their characteristics. During the productive life of a reservoir, its corresponding simulation model needs to be adjusted periodically because the availability of an appropriate model is crucial to obtain accurate predictions about the production, and this directly impacts the management decisions. The adjustment of the properties of the model is translated into a complex optimization problem, where the number of variables involved increases with the increase of the number of blocks that make up the mesh of the simulation model, requiring too much effort on the part of a specialist. The availability of a computational tool that can assist the specialist on part of this process can be very useful both for obtaining quicker responses, as for making better decisions. Thus, this work combines computational intelligence through Coevolutionary Genetic Algorithm with Multipoint Geostatistics, proposing and implementing an architecture optimization applied to the tuning properties of reservoir models. This architecture differs from traditional approaches to be able to optimize simultaneously more than one property of the reservoir simulation model. We used also distributed processing to explore the parallel computing power of genetic algorithms. The architecture was capable of generating models that adequately fit the curves of production, preserving the consistency and continuity of the geological reservoir obtaining, respectively, 98 percent and 97 percent of reduction in error of fit to the historical data and forecasting. For porosity and permeability maps, the reductions in errors were 79 percent and 84 percent, respectively.

Kartézské genetické programován (CGP) je evoluc inspirovaná metoda strojového učen, která je primárně určená pro automatizovaný návrh programů a čslicových obvodů. CGP je úspěšné v řešen mnoha úloh z reálného světa. Avšak k nalezen inovativnch řešen obvykle potřebuje značný výpočetn výkon. Každý kandidátn program navržený pomoc CGP mus být spuštěn, aby se zjistilo, do jaké mry tento program řeš zadaný problém, a mohla mu být přiřazena fitness hodnota. Právě vyhodnocen fitness bývá výpočetně nejnáročnějš část návrhu pomoc CGP. Tato práce se zabývá využitm koevoluce prediktorů fitness v CGP za účelem zrychlen procesu evolučnho návrhu prováděného pomoc CGP. Prediktor fitness je malá podmnožina trénovacch dat použvaná pro rychlý odhad fitness hodnoty namsto náročného vyhodnocen objektivn fitness hodnoty. Koevoluce prediktorů fitness je optimalizačn metoda modelován fitness, která snižuje náročnost a frekvenci výpočtu fitness. V této práci je koevolučn algoritmus přizpůsoben pro CGP a jsou představeny a zkoumány tři přstupy k zakódován prediktorů fitness. Představená metoda je experimentálně vyhodnocena v pěti úlohách symbolické regrese a v úloze návrhu obrazových filtrů. Výsledky experimentů ukazuj, že pomoc této metody lze významně snžit výpočetn čas, který CGP potřebuje pro řešen zkoumané třdy úloh.

This thesis deals with image filter design using coevolutionary algorithms. It contains a description of evolutionary algorithms, focusing on genetic programming, cartesian genetic programming and coevolution, the reader can learn about image filters too. The next chapters contain the design of image filters and noise detectors using cooperative coevolution, and the implementation and testing of the proposed filter. In the last chapter the proposed filter is compared to other filters created using evolutionary algorithms but without coevolution.

This thesis deals with the usage of coevolution in the task of symbolic regression. Symbolic regression is used for obtaining mathematical formula which approximates the measured data. It can be executed by genetic programming - a method from the category of evolutionary algorithms that is inspired by natural evolutionary processes. Coevolution works with multiple evolutionary processes that are running simultaneously and influencing each other. This work deals with the design and implementation of the application which performs symbolic regression using coevolution on test-based problems. The test set was generated by a new method, which allows to adjust its size dynamically. Functionality of the application was verified on a set of five test tasks. The results were compared with a coevolution algorithm with a fixed-sized test set. In three cases the new method needed lesser number of generations to find a solution of a desired quality, however, in most cases more data-point evaluations were required.

No description available.

This thesis deals with employing coevolutionary principles to the image filter design. Evolutionary algorithms are very advisable method for image filter design. Using coevolution, we can add the processes, which can accelerate the convergence by interactions of candidate filters population with population of fitness predictors. Fitness predictor is a small subset of the training set and it is used to approximate the fitness of the candidate solutions. In this thesis, indirect encoding is used for predictors evolution. This encoding represents a mathematical expression, which selects training vectors for candidate filters fitness prediction. This approach was experimentally evaluated in the task of image filters for various intensity of random impulse and salt and pepper noise design and the design of the edge detectors. It was shown, that this approach leads to adapting the number of target objective vectors for a particular task, which leads to computational complexity reduction.

Submitted by Jaqueline Silva (jtas29@gmail.com) on 2014-10-06T17:24:11Z No. of bitstreams: 2 Dissertação - André Assis Lôbo de Oliveira- 2013.pdf: 3915731 bytes, checksum: 2447fa437e5dca74e295727bd8fed4d1 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) / Approved for entry into archive by Jaqueline Silva (jtas29@gmail.com) on 2014-10-06T19:18:43Z (GMT) No. of bitstreams: 2 Dissertação - André Assis Lôbo de Oliveira- 2013.pdf: 3915731 bytes, checksum: 2447fa437e5dca74e295727bd8fed4d1 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) / Made available in DSpace on 2014-10-06T19:18:43Z (GMT). No. of bitstreams: 2 Dissertação - André Assis Lôbo de Oliveira- 2013.pdf: 3915731 bytes, checksum: 2447fa437e5dca74e295727bd8fed4d1 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) Previous issue date: 2013-12-05 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / Verification and Validation Activities (V&V) consume about 50% to 60% of the total cost of a software lifecycle. Among those activities, Software Testing technique is one which is mostly used during this process. One of the main problems related to detected in Software Testing is to find a set of tests (subset from input domain of the problem) which is effective to detect the remaining bugs in the software. The Search-Based Software Testing (SBST) approach uses metaheuristics to find low cost set of tests with a high effectiveness to detect bugs. From several existing test criteria, Mutation Testing is considered quite promising to reveal bugs, despite its high computational cost, due to the great quantity of mutant programs generated. Therefore, this dissertation addresses the problem of selecting mutant programs and test cases in Mutation Testing context. To this end, it is proposed a Coevolutionary Genetic Algorithm (CGA) and the concept of Genetic Effectiveness, implemented by Genetic Classification (GC) and new genetic operators adapted to the proposed representation. Furthermore, the Genetic Algorithm Coevolutionary with Controlled Genetic Classification (CGA􀀀CGCop) is proposed for improving the efficiency of CGA’s GC. The CGA is applied in three categories of benchmarks and compared to other five methods. The results show a better performance of the CGA in subsets selection with better mutation score, as well as improvement of CGA􀀀CGCop in use of GC. These results evidence the proposal approach with promising use in the context of Mutation Testing. / Atividades de Validação e Verificação (V&V) consomem cerca de 50% a 60% do custo total no ciclo de vida de um software. Dentre essas, o Teste de Software é uma das atividades mais empregadas. Um dos maiores problemas do Teste de Software é encontrar um conjunto de teste (subconjunto do domínio de entrada do problema) que seja eficaz em detectar os defeitos remanescentes no software. Neste contexto, a Search-Based Software Testing (SBST) é uma linha de pesquisa recente que vem propondo boas soluções, uma vez que utiliza-se de metaheurísticas para encontrar um conjunto de teste com baixo custo e grande eficácia na detecção de defeitos. Dentre os diversos critérios de teste existentes, o Teste de Mutação é bastante promissor na revelação de defeitos, entretanto apresenta um alto custo computacional em termos de aplicabilidade. Por isso, a pesquisa aborda o problema de seleção de programas mutantes e casos de teste no contexto do Teste de Mutação. Para tal, propõe o Algoritmo Genético Coevolucionário (AGC) que traz o conceito de Efetividade Genética, implementado pela Classificação Genética (CG) e por novos operadores genéticos adaptados à representação proposta. Além disso, propõe o Algoritmo Genético Coevolucionário com Classificação Genética Controlada (AGC 􀀀CGCop) para a melhoria da eficiência da CG do AGC. O algoritmo AGC é aplicado em três classes de benchmarks e comparado com outros cinco métodos. Os resultados demonstram um melhor desempenho do AGC na seleção de subconjuntos com melhor escore de mutação, bem como um aprimoramento do AGC􀀀CGCop no uso da CG. Tais resultados evidenciam a abordagem proposta com uso promissor no contexto do Teste de Mutação.

The aim of this work is to automatically design a program that is able to detect dyskinetic movement features in the measured patient's movement data. The program will be developed using Cartesian genetic programming equipped with coevolution of fitness predictors. This type of coevolution allows to speed up a design performed by Cartesian genetic programming by evaluating a quality of candidate solutions using only a part of training data. Evolved classifier achieves a performance (in terms of AUC) that is comparable with the existing solution while achieving threefold acceleration of the learning process compared to the variant without the fitness predictors, in average. Experiments with crossover methods for fitness predictors haven't shown a significant difference between investigated methods. However, interesting results were obtained while investigating integer data types that are more suitable for implementation in hardware. Using an unsigned eight-bit data type (uint8_t) we've achieved not only comparable classification performance (for significant dyskinesia AUC = 0.93 the same as for the existing solutions), with improved AUC for walking patient's data (AUC = 0.80, while existing solutions AUC = 0.73), but also nine times speedup of the design process compared to the approach without fitness predictors employing the float data type, in average.