Satellite imagery and remote sensing provide explanatory variables at relatively high resolutions for modeling geospatial phenomena, yet regional summaries are often desirable for analysis and actionable insight. In this paper, we propose a novel method of inducing spatial aggregations as a component of the statistical learning process, yielding regional model features whose construction is driven by model prediction performance rather than prior assumptions. Our results demonstrate that Genetic Programming is particularly well suited to this type of feature construction because it can automatically synthesize appropriate aggregations, as well as better incorporate them into predictive models compared to other regression methods we tested. In our experiments we consider a specific problem instance and real-world dataset relevant to predicting snow properties in high-mountain Asia.

Tree-based genetic programming (GP) has several known shortcomings: difficult adaptability to specific programming languages and environments, the problem of closure and multiple types, and the problem of declarative representation of knowledge. Most of the methods that try to solve these problems are based on formal grammars. The precise effect of their distinctive features is often difficult to analyse and a good comparison of performance in specific problems is missing. This thesis reviews three grammar-based methods: context-free grammar genetic programming (CFG-GP), including its variant GPHH recently applied to exam timetabling, grammatical evolution (GE), and LOGENPRO, it discusses how they solve the problems encountered by GP, and compares them in a series of experiments in six applications using success rates and derivation tree characteristics. The thesis demonstrates that neither GE nor LOGENPRO provide a substantial advantage over CFG-GP in any of the experiments, and analyses the differences between the effects of operators used in CFG-GP and GE. It also presents results from a highly efficient implementation of CFG-GP and GE.

This thesis deals with a problem of choosing the most suitable agent for a new data mining task not yet seen by the agents. The metric is proposed on the data mining tasks space, and based on this metric similar tasks are identified. This set is advanced as an input to a program evolved by means of genetic programming. The program estimates agents performance on the new task from both the time and error point of view. A JADE agent is implemented which provides an interface allowing other agents to obtain estimation results in real time.

[pt] A superioridade de desempenho dos algoritmos quânticos, em alguns problemas específicos, reside no uso direto de fenômenos da mecânica quântica para realizar operações com dados em computadores quânticos. Esta característica fez surgir uma nova abordagem, denominada Computação com Inspiração Quântica, cujo objetivo é criar algoritmos clássicos (executados em computadores clássicos) que tirem proveito de princípios da mecânica quântica para melhorar seu desempenho. Neste sentido, alguns algoritmos evolutivos com inspiração quântica tem sido propostos e aplicados com sucesso em problemas de otimização combinatória e numérica, apresentando desempenho superior àquele dos algoritmos evolutivos convencionais, quanto à melhoria da qualidade das soluções e à redução do número de avaliações necessárias para alcançá-las. Até o presente momento, no entanto, este novo paradigma de inspiração quântica ainda não havia sido aplicado à Programação Genética (PG), uma classe de algoritmos evolutivos que visa à síntese automática de programas de computador. Esta tese propõe, desenvolve e testa um novo modelo de algoritmo evolutivo com inspiração quântica, denominado Programação Genética Linear com Inspiração Quântica (PGLIQ), para a evolução de programas em código de máquina. A Programação Genética Linear é assim denominada porque cada um dos seus indivíduos é representado por uma lista de instruções (estruturas lineares), as quais são executadas sequencialmente. As contribuições deste trabalho são o estudo e a formulação inédita do uso do paradigma da inspiração quântica na síntese evolutiva de programas de computador. Uma das motivações para a opção pela evolução de programas em código de máquina é que esta é a abordagem de PG que, por oferecer a maior velocidade de execução, viabiliza experimentos em larga escala. O modelo proposto é inspirado em sistemas quânticos multiníveis e utiliza o qudit como unidade básica de informação quântica, o qual representa a superposição dos estados de um sistema deste tipo. O funcionamento do modelo se baseia em indivíduos quânticos, que representam a superposição de todos os programas do espaço de busca, cuja observação gera indivíduos clássicos e os programas (soluções). Nos testes são utilizados problemas de regressão simbólica e de classificação binária para se avaliar o desempenho da PGLIQ e compará-lo com o do modelo AIMGP (Automatic Induction of Machine Code by Genetic Programming), considerado atualmente o modelo de PG mais eficiente na evolução de código de máquina, conforme citado em inúmeras referências bibliográficas na área. Os resultados mostram que a Programação Genética Linear com Inspiração Quântica (PGLIQ) apresenta desempenho geral superior nestas classes de problemas, ao encontrar melhores soluções (menores erros) a partir de um número menor de avaliações, com a vantagem adicional de utilizar um número menor de parâmetros e operadores que o modelo de referência. Nos testes comparativos, o modelo mostra desempenho médio superior ao do modelo de referência para todos os estudos de caso, obtendo erros de 3 a 31% menores nos problemas de regressão simbólica, e de 36 a 39% nos problemas de classificação binária. Esta pesquisa conclui que o paradigma da inspiração quântica pode ser uma abordagem competitiva para se evoluir programas eficientemente, encorajando o aprimoramento e a extensão do modelo aqui apresentado, assim como a criação de outros modelos de programação genética com inspiração quântica. / [en] The superior performance of quantum algorithms in some specific problems lies in the direct use of quantum mechanics phenomena to perform operations with data on quantum computers. This feature has originated a new approach, named Quantum-Inspired Computing, whose goal is to create classic algorithms (running on classical computers) that take advantage of quantum mechanics principles to improve their performance. In this sense, some quantum-inspired evolutionary algorithms have been proposed and successfully applied in combinatorial and numerical optimization problems, presenting a superior performance to that of conventional evolutionary algorithms, by improving the quality of solutions and reducing the number of evaluations needed to achieve them. To date, however, this new paradigm of quantum inspiration had not yet been applied to Genetic Programming (GP), a class of evolutionary algorithms that aims the automatic synthesis of computer programs. This thesis proposes, develops and tests a novel model of quantum-inspired evolutionary algorithm named Quantum-Inspired Linear Genetic Programming (QILGP) for the evolution of machine code programs. Linear Genetic Programming is so named because each of its individuals is represented by a list of instructions (linear structures), which are sequentially executed. The contributions of this work are the study and formulation of the novel use of quantum inspiration paradigm on evolutionary synthesis of computer programs. One of the motivations for choosing by the evolution of machine code programs is because this is the GP approach that, by offering the highest speed of execution, makes feasible large-scale experiments. The proposed model is inspired on multi-level quantum systems and uses the qudit as the basic unit of quantum information, which represents the superposition of states of such a system. The model’s operation is based on quantum individuals, which represent a superposition of all programs of the search space, whose observation leads to classical individuals and programs (solutions). The tests use symbolic regression and binary classification problems to evaluate the performance of QILGP and compare it with the AIMGP model (Automatic Induction of Machine Code by Genetic Programming), which is currently considered the most efficient GP model to evolve machine code, as cited in numerous references in this field. The results show that Quantum-Inspired Linear Genetic Programming (QILGP) presents superior overall performance in these classes of problems, by achieving better solutions (smallest error) from a smaller number of evaluations, with the additional advantage of using a smaller number of parameters and operators that the reference model. In comparative tests, the model shows average performance higher than that of the reference model for all case studies, achieving errors 3-31% lower in the problems of symbolic regression, and 36-39% in the binary classification problems. This research concludes that the quantum inspiration paradigm can be a competitive approach to efficiently evolve programs, encouraging the improvement and extension of the model presented here, as well as the creation of other models of quantum-inspired genetic programming.

A morfologia matemática fornece ferramentas poderosas para a realização de análise de imagens em baixo nível e tem encontrado aplicações em diversas áreas, tais como: visão robótica, inspeção visual, medicina, análise de textura, entre outras. Muitas dessas aplicações requerem processamento em tempo real e para sua execução de forma eficiente freqüentemente é utilizado hardware dedicado. Também, a tarefa de projetar operadores morfológicos manualmente para uma dada aplicação não é trivial na prática. A programação genética, que é um ramo relativamente novo em computação evolucionária, está se consolidando como um método promissor em aplicações envolvendo processamento de imagens digitais. Seu objetivo primordial é descobrir como os computadores podem aprender a resolver problemas sem, no entanto, serem programados para essa tarefa. Essa área ainda não foi muito explorada no contexto de construção automática de operadores morfológicos. Assim, neste trabalho, desenvolve-se e implementa-se uma arquitetura original, de baixo custo, reconfigurável por meio de instruções morfológicas e lógicas geradas automaticamente através de uma aproximação linear baseada em programação genética, visando-se o processamento morfológico de imagens em tempo real utilizando FPGAs de alta complexidade, com objetivos de filtragem, reconhecimento de padrões e emulação de filtros desconhecidos de softwares comerciais, para citar somente algumas aplicações. Exemplos de aplicações práticas envolvendo imagens binárias, em níveis de cinza e coloridas são fornecidos e seus resultados são comparados com outras formas de implementação. / Mathematical morphology supplies powerful tools for low level image analysis, with applications in robotic vision, visual inspection, medicine, texture analysis and many other areas. Many of the mentioned applications require dedicated hardware for real time execution. The task of designing manually morphological operators for a given application isnot always a trivial one. Genetic programming is a relatively new branch of evolutionary computing and it is consolidating as a promising method for applications of digital image processing. The main objective of genetic programming is to discover how computers can learn to solve problems without being programmed for that. In the literature little has been found about the automatic morphological operators construction using genetic programming. In this work, the development of an original reconfigurable architecture using logical and morphological instructions generated automatically by a linear approach based on genetic programming is presented. The developed architecture is based on Field Programmable Gate Arrays (FPGAs) and has among the possible applications, image filtering, pattern recognition and filter emulation. Binary, gray level and color image practical applications using the developed architecture are presented and the results are compared with other implementation techniques.

Lau, Wai Shing. / Thesis submitted in: November 2006. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2007. / Includes bibliographical references (leaves 85-94). / Abstracts in English and Chinese. / Abstract --- p.i / Acknowledgement --- p.iv / Chapter 1 --- Introduction --- p.1 / Chapter 1.1 --- Field Programmable Gate Arrays --- p.2 / Chapter 1.2 --- FPGA technology mapping problem --- p.3 / Chapter 1.3 --- Motivations --- p.5 / Chapter 1.4 --- Contributions --- p.6 / Chapter 1.5 --- Thesis Organization --- p.9 / Chapter 2 --- Background Study --- p.11 / Chapter 2.1 --- Deterministic approach to technology mapping problem --- p.11 / Chapter 2.1.1 --- FlowMap --- p.12 / Chapter 2.1.2 --- DAOMap --- p.14 / Chapter 2.2 --- Stochastic approach --- p.15 / Chapter 2.2.1 --- Bio-Inspired Methods for Multi-Level Combinational Logic Circuit Design --- p.15 / Chapter 2.2.2 --- A Survey of Combinational Logic Circuit Representations in stochastic algorithms --- p.17 / Chapter 2.3 --- Genetic Parallel Programming --- p.20 / Chapter 2.3.1 --- Accelerating Phenomenon --- p.22 / Chapter 2.4 --- Chapter Summary --- p.23 / Chapter 3 --- A GPP based Logic Circuit Synthesizer --- p.24 / Chapter 3.1 --- Overall system architecture --- p.25 / Chapter 3.2 --- Multi-Logic-Unit Processor --- p.26 / Chapter 3.3 --- The Genotype of a MLP program --- p.28 / Chapter 3.4 --- The Phenotype of a MLP program --- p.31 / Chapter 3.5 --- The Evolution Engine --- p.33 / Chapter 3.5.1 --- The Dual-Phase Approach --- p.33 / Chapter 3.5.2 --- Genetic operators --- p.35 / Chapter 3.6 --- Chapter Summary --- p.38 / Chapter 4 --- MLP in hardware --- p.39 / Chapter 4.1 --- Motivation --- p.39 / Chapter 4.2 --- Hardware Design and Implementation --- p.40 / Chapter 4.3 --- Experimental Settings --- p.43 / Chapter 4.4 --- Experimental Results and Evaluations --- p.46 / Chapter 4.5 --- Chapter Summary --- p.50 / Chapter 5 --- Feasibility Study of Multi MLPs --- p.51 / Chapter 5.1 --- Motivation --- p.52 / Chapter 5.2 --- Overall Architecture --- p.53 / Chapter 5.3 --- Experimental settings --- p.55 / Chapter 5.4 --- Experimental results and evaluations --- p.59 / Chapter 5.5 --- Chapter Summary --- p.59 / Chapter 6 --- A Hybridized GPPLCS --- p.61 / Chapter 6.1 --- Motivation --- p.62 / Chapter 6.2 --- Overall system architecture --- p.62 / Chapter 6.3 --- Experimental settings --- p.64 / Chapter 6.4 --- Experimental results and evaluations --- p.66 / Chapter 6.5 --- Chapter Summary --- p.70 / Chapter 7 --- A Memetic GPPLCS --- p.71 / Chapter 7.1 --- Motivation --- p.72 / Chapter 7.2 --- Overall system architecture --- p.72 / Chapter 7.3 --- Experimental settings --- p.76 / Chapter 7.4 --- Experimental results and evaluations --- p.77 / Chapter 7.5 --- Chapter Summary --- p.80 / Chapter 8 --- Conclusion --- p.82 / Chapter 8.1 --- Future work --- p.83 / Bibliography --- p.85

Digital games are part of our culture and have gained significant attention over the last decade. The growing capabilities of home computers, gaming consoles and mobile phones allow current games to visualise 3D virtual worlds, photo-realistic characters and the inclusion of complex physical simulations. The growing computational power of those devices enables the usage of complex algorithms while visualising data. Therefore, opportunities arise for developers of interactive products such as digital games which introduce new, challenging and exciting elements to the next generation of highly interactive software systems. Two of those challenges, which current systems do not address adequately, are design support for creating Intelligent Virtual Agents and more believable non-player characters for immersive game-play. We start in this thesis by addressing the agent design support first and then extend the research, addressing the second challenge. The main contributions of this thesis are: - The POSH-SHARP system is a framework for the development of game agents. The platform is modular, extendable, offers multi-platform support and advanced software development features such as behaviour inspection and behaviour versioning. The framework additionally integrates an advanced information exchange mechanism supporting loose behaviour coupling. - The Agile behaviour design methodology integrates agile software development and agent design. To guide users, the approach presents a work-flow for agent design and guiding heuristics for their development. - The action selection augmentation ERGo introduces a "white-box" solution to altering existing agent frameworks, making their agents less deterministic. It augments selected behaviours with a bio-mimetic memory to track and adjust their activation over time. With the new approach to agent design, the development of "deepagent" behaviour for digital adversaries and advanced tools supporting their design is given. Such mechanisms should enable developers to build robust non-player characters that act more human-like in an efficient and robust manner. Within this thesis, different strategies are identified to support the design of agents in a more robust manner and to guide developers. These discussed mechanisms are then evolved to develop and design Intelligent Virtual Agents. Because humans are still the best measurement for human-likeness, the evolutionary cycle involves feedback given by human players.

Le but de la thèse est de développer des stratégies de contrôle efficaces pour la réduction de la trainée aérodynamique des véhicules terrestres. Nous examinons expérimentalement les effets d’un forçage fluidique sur le sillage d’un modèle de véhicule simplifié à culot droit. Le forçage est effectué par des jets pulsés aux arêtes et16 capteurs de pression répartis à la surface arrière permettent d’estimer la traînée instantanée. Nous abordons le problème difficile du contrôle de l’écoulement turbulent non linéaire---qui est souvent au-delà des capacités de la modélisation réduite---par le développement d'une stratégie de contrôle sans modèle: le contrôle via la programmation génétique linéaire (LGPC) dirigé par les données. Cette méthode explore et exploite la dynamique fortement non linéaire d'une manière non supervisée avec pas ou peu de connaissances antérieures sur le système.Le problème est de trouver une logique de contrôle qui optimise une fonction de coût donnée. Cette optimisation est réalisée par la programmation génétique linéaire comme un solveur de régression simple dans un espace de recherche de grande dimension. En particulier, cette recherche fait progresser et généralise les études antérieures sur le contrôle via la programmation génétique en incluant le forçage multi-fréquences, le signal des capteurs,l’historique des informations temporelles et leurs combinaisons dans l'espace de recherche de contrôle. La performance de LGPC est démontrée avec succès sur les expériences de contrôle de traînée du modèle de véhicule simplifié où le sillage turbulent présente une symétrie latérale et une asymétrie normale à la paroi. Environ 33% de récupération de pression au culot associée à 22% de réduction de trainée est obtenue dans toutes les classes de loisde contrôle considérées. L'énergie consommée du forçage ne représente que 30% de l'énergie aérodynamique récupérée. Dans ce travail, nous étudions également les sillages turbulents ayant une asymétrie latérale: un sillage intermittent et bi-modal à dérapage nul et un sillage asymétrique avec un angle de dérapage modéré de 5 degrés.Pour le sillage intermittent, un contrôle de rétroaction en opposition basé sur la physique est déduit à partir des essais précédents de contrôle en boucle ouverte. Le contrôleur supprime avec succès la bi-modalité du sillage et rend le sillage symétrique avec une réduction de traînée concomitante. Pour le sillage asymétrique en dérapage,nous construisons un contrôle bi-fréquence à l’arête au vent à partir des résultats de forçage à fréquence unique. Ce forçage bi-fréquentiel comprend deux fréquences ayant une différence d'un ordre de grandeur. Il combine les effets favorables de la vectorisation du sillage et le contrôle de l'équilibre des couches de cisaillement. Il est important de noter que la stratégie LGPC est également appliqué à cette situation en dérapage et converge vers le même forçage bi-fréquentiel. Les stratégies de contrôle proposées dans cette étude ouvrent de nouveaux chemins prometteurs pour le contrôle de la réduction de la traînée dans des conditions plus complexes de vitesse amont variable ou de rafale. / The thesis aims to develop effective active flow control strategies for aerodynamic drag reduction of road vehicles.We experimentally examine the effects of fluidic actuation on the wake past a simplified square-back car model.The actuation is performed with pulsed jets at trailing edges and the flow is monitored with 16 pressure sensors distributed at the rear side. We address the challenging nonlinear turbulence control---which is often beyond the capabilities of model-oriented approach---by developing a simple yet powerful model-free control strategy: the data-driven linear genetic programming control (LGPC). This method explores and exploits strongly nonlinear dynamics in an unsupervised manner with no or little prior knowledge about the system. The control problem is to find a control logic which optimizes a given cost function by employing linear genetic programming as an easy and simple regression solver in a high-dimensional control search space. In particular, the present work advances and generalizes the previous studies of genetic programming control by comprising multi-frequency forcing, sensor-based feedback including also time-history information feedback and combinations thereof in the control search space. The performance of LGPC is successfully demonstrated on the drag control experiments of the car model where the investigated turbulent wake exhibits a spanwise symmetry and a wall-normal asymmetry. Approximately 33% base pressure recovery associated with 22% drag reduction is achieved in all considered classes of control laws. The consumed actuation energy accounts for only 30% of the aerodynamic power saving. In this research, we also study the turbulent wakes having a lateral asymmetry: an intermittent bi-modal wake at zero yaw and an asymmetric wake at a moderate yaw angle of 5 degree. For the bimodal wake exhibiting are flectional symmetry-breaking, a physics-based opposition feedback control is inferred from the previous open loop control tests. The controller successfully suppresses the bi-modality of the wake and renders a symmetrized wake with a concomitant drag reduction. For the asymmetric wake at yaw, we infer from the single-frequency forcing results a bi-frequency control at the windward edge comprising two frequencies having one order of magnitude difference. This bi-frequency actuation combines the favorable effects of fluidic boat-tailing and balance control of the shear layers. Importantly, LGPC is also applied to this yawed situation and converges to the same bi-frequency actuation. The control strategies proposed in the present study open promising new paths for the control of drag reduction in more complex conditions such as the varying oncoming velocity and wind gust.

This thesis deals with trading strategy development on foreign market exchange (FOREX) from a perspective of technical analysis. A hypothesis that price data can be used for stable profitable decision making is analyzed. For that purpose, financial data preparation and derived indicators are described in detail. At first, strategies are randomly constructed. Afterward, they are improved iteratively by means of evolution principles. Genetic programming is used in particular. Special attention is devoted to fitness functon definition, on which the progress of strategies depends. Besides usual criterion - profit, other aspects are taken into account. Transactional costs are included in strategy evaluation. Costs have even been increased in comparison with declared values. Some of the best strategies are tested on data, which has not been used for their development. Obtained results did not support the hypothesis. However, process of a strategy search can be improved.

Web Information Extraction (WIE) is the discipline dealing with the discovery, processing and extraction of specific pieces of information from semi-structured or unstructured web pages. The World Wide Web comprises billions of web pages and there is much need for systems that will locate, extract and integrate the acquired knowledge into organisations practices. There are some commercial, automated web extraction software packages, however their success comes from heavily involving their users in the process of finding the relevant web pages, preparing the system to recognise items of interest on these pages and manually dealing with the evaluation and storage of the extracted results. This research has explored WIE, specifically with regard to the automation of the extraction and validation of online training information. The work also includes research and development in the area of automated Web Information Retrieval (WIR), more specifically in Web Searching (or Crawling) and Web Classification. Different technologies were considered, however after much consideration, Naïve Bayes Networks were chosen as the most suitable for the development of the classification system. The extraction part of the system used Genetic Programming (GP) for the generation of web extraction solutions. Specifically, GP was used to evolve Regular Expressions, which were then used to extract specific training course information from the web such as: course names, prices, dates and locations. The experimental results indicate that all three aspects of this research perform very well, with the Web Crawler outperforming existing crawling systems, the Web Classifier performing with an accuracy of over 95% and a precision of over 98%, and the Web Extractor achieving an accuracy of over 94% for the extraction of course titles and an accuracy of just under 67% for the extraction of other course attributes such as dates, prices and locations. Furthermore, the overall work is of great significance to the sponsoring company, as it simplifies and improves the existing time-consuming, labour-intensive and error-prone manual techniques, as will be discussed in this thesis. The prototype developed in this research works in the background and requires very little, often no, human assistance.