There is a continuing demand for higher performance, particularly in the area of scientific and engineering computation. In order to achieve high performance in the context of frequent hardware upgrading, software must be adaptable for portable performance. What is required is an optimising compiler that evolves and adapts itself to environmental change without sacrificing performance. Java has emerged as a dominant programming language widely used in a variety of application areas. However, its architectural independant design means that it is frequently unable to deliver high performance especially when compared to other imperative languages such as Fortran and C/C++. This thesis presents a language- and architecture-independant approach to achieve portable high performance. It uses the mapping notation introduced in the Unified Transformation Framework to specify a large optimisation space. A heuristic random search algorithm is introduced to explore this space in a feedback-directed iterative optimisation manner. It is then extended using a machine learning approach which enables the compiler to learn from its previous optimisations and apply the knowledge when necessary. Both the heuristic random search algorithm and the learning optimisation approach are implemented in a prototype Adaptive Optimisation Framework for Java (AOF-Java). The experimental results show that the heuristic random search algorithm can find, within a relatively small number of atttempts, good points in the large optimisation space. In addition, the learning optimisation approach is capable of finding good transformations for a given program from its prior experience with other programs.

The container loading problem considered in this thesis is to determine placements of a set of packages within one or multiple shipping containers. Smaller packages are consolidated on pallets prior to being loaded in the shipping containers together with larger packages. There are multiple objectives which may be summarized as fitting all the packages while achieving good stability of the cargo as well as the shipping containers themselves. According to recent literature reviews, previous research in the field have to large extent been neglecting issues relevant in practice. Our real-world application was developed for the industrial company Atlas Copco to be used for sea container shipments at their Distribution Center (DC) in Texas, USA. Hence all applicable practical constraints faced by the DC operators had to be treated properly. A high variety in sizes, weights and other attributes such as stackability among packages added complexity to an already challenging combinatorial problem. Inspired by how the DC operators plan and perform loading manually, the batch concept was developed, which refers to grouping of boxes based on their characteristics and solving subproblems in terms of partial load plans. In each batch, an extensive placement heuristic and a load plan evaluation run iteratively, guided by a Genetic Algorithm (GA). In the placement heuristic, potential placements are evaluated using a scoring function considering aspects of the current situation, such as space utilization, horizontal support and heavier boxes closer to the floor. The scoring function is weighted by coefficients corresponding to the chromosomes of an individual in the GA population. Consequently, the fitness value of an individual in the GA population is the rating of a load plan. The loading optimization software has been tested and successfully implemented at the DC in Texas. The software has been proven capable of generating satisfactory load plans within acceptable computation times, which has resulted in reduced uncertainty and labor usage in the loading process. Analysis using real sea container shipments shows that the GA is able to tune the scoring coefficients to suit the particular problem instance being solved.

Well-publicized lost opportunities for U.S. and coalition air forces to strike enemy leadership targets in Afghanistan and Iraq demonstrate the importance of Time Sensitive Targeting. How do we "pair" the weapon and weapons delivery platform with their target? The available platforms (aircraft, manned or unmanned) may be on the ground in an alert status, loitering airborne, or on their way to attack other targets. The problem is compounded by the facts that we actually wish to (a) create multiple strike packages simultaneously, (b) recompose existing strike packages that are disrupted by the new plans, (c) minimize such disruptions, (d) satisfy minimum kill probabilities, and (e) avoid the attrition of tasked assets. This thesis develops an automated, optimizing, heuristic decision aid, "RAPT-OR", that rapidly revises a current Air Taking Order (ATO) to meet the requirements above. Using a set-packing model, RAPT-OR, an ATO near optimally, on a desktop PC, in less than two seconds, for a typical scenario with 40 aircraft, four new targets and hundreds of potential strike packages. RAPT-OR allows decision makers the ability of adjusting risk acceptance in the formulation of possible courses of action by manipulating friendly attrition importance in formulating a solution.

Mémoire numérisé par la Division de la gestion de documents et des archives de l'Université de Montréal.

Data in real-world graph drawing applications often change frequently but incrementally. Any drastic change in the graph layout could disrupt a user's "mental map." Furthermore, real-world applications like enterprise process or e-commerce graphing, where data change rapidly in both content and quantity, demand a comprehensive responsiveness when rendering the graph layout in a multi-user environment in real time. Most standard static graph drawing algorithms apply global changes and redraw the entire graph layout whenever the data change. The new layout may be very different from the previous layout and the time taken to redraw the entire graph degrades quickly as the amount of graph data grows. Dynamic behavior and the quantity of data generated by real-world applications pose challenges for existing graph drawing algorithms in terms of incremental stability and scalability. A constrained hierarchical graph drawing framework and modified Sugiyama heuristic were developed in this research. The goal of this research was to improve the scalability of the constrained graph drawing framework while preserving layout stability. The framework's use of the relational data model shifts the graph application from the traditional desktop to a collaborative and distributed environment by reusing vertex and edge information stored in a relational database. This research was based on the work of North and Woodhull (2001) and the constrained crossing reduction problem proposed by Forster (2004). The result of the constrained hierarchical graph drawing framework and the new Sugiyama heuristic, especially the modified barycenter algorithms, were tested and evaluated against the Graphviz framework and North and Woodhull's (2001) online graph drawing framework. The performance test results showed that the constrained graph drawing framework run time is comparable with the performance of the Graphviz framework in terms of generating static graph layouts, which is independent of database accesses. Decoupling graph visualization from the graph editing modules improved scalability, enabling the rendering of large graphs in real time. The visualization test also showed that the constrained framework satisfied the aesthetic criteria for constrained graph layouts. Future enhancements for this proposed framework include implementation of (1) the horizontal coordinate assignment algorithm, (2) drawing polylines for multilayer edges in the rendering module, and (3) displaying subgraphs for very large graph layouts.

Today, there is a substantial number of software and research groups that focus on the development of image processing software to extract useful information from medical images, in order to assist and improve patient diagnosis. The work presented in this thesis is centred on processing of images of blood and bone marrow smears of patients suffering from leukaemia, a common type of cancer. In general, cancer is due to aberrant gene expression, which is caused by either mutations or epigenetic changes in DNA. Poor diet and unhealthy lifestyle may trigger or contribute to these changes, although the underlying mechanism is often unknown. Importantly, many cancer types including leukaemia are curable and patient survival and treatment can be improved, subject to prompt diagnosis. In particular, this study focuses on Acute Myeloid Leukaemia (AML), which can be of eight distinct types (M0 to M7), with the main objective to develop a methodology to automatically detect and classify leukaemia cells into one of the above types. The data was collected from the Department of Haematology, Universiti Sains Malaysia, in Malaysia. Three main methods, namely Cellular Automata, Heuristic Search and classification using Neural Networks are facilitated. In the case of Cellular Automata, an improved method based on the 8-neighbourhood and rules were developed to remove noise from images and estimate the radius of the potential blast cells contained in them. The proposed methodology selects the starting points, corresponding to potential blast cells, for the subsequent seeded heuristic search. The Seeded Heuristic employs a new fitness function for blast cell detection. Furthermore, the WEKA software is utilised for classification of blast cells and hence images, into AML subtypes. As a result accuracy of 97.22% was achieved in the classification of blasts into M3 and other AML subtypes. Finally, these algorithms are integrated into an automated system for image processing. In brief, the research presented in this thesis involves the use of advanced computational techniques for processing and classification of medical images, that is, images of blood samples from patients suffering from leukaemia.

Crew schedules on the South African Transport Services are done by roster compilers at depots. A method that uses heuristic and mathematical programming algorithms was developed to replace existing hand methods. It is a two stage method that will use a microcomputer to assist roster compilers to draw up crew schedules. Initially timetables are subdivided into shifts and then they are combined into crew schedules. The solution, which produces a significant improvement compared with an existing crew schedule and an existing method, has been accepted in principle and computer programming has begun. In Appendix E another heuristic for the scheduling of league matches is described.

Sistemas de atendimento emergencial estão presentes nos grandes centros urbanos de todo o mundo. Sua finalidade é fornecer socorro a indivíduos acidentados, catástrofes e pessoas com problemas de saúde de maneira ágil e eficaz. Assim, uma característica importante desses sistemas é a prontidão dos recursos e o planejamento da malha de atendimento, definida pelas localizações das bases de veículos e pelas viaturas de atendimento. A operação desses sistemas é ainda mais crítica em grandes centros urbanos, nos quais as condições de trânsito e os padrões de variação da demanda por atendimento emergencial inserem um grau de complexidade extra ao problema. Desta forma, um mapa de localização de bases e viaturas eficientemente elaborado é crítico para o desempenho do sistema como um todo. Como no Brasil existe uma lacuna entre os mecanismos de planejamento dessas malhas de atendimento e as teorias de localização de instalações, observa-se que os métodos atualmente utilizados para a localização das bases e veículos, bem como para o dimensionamento desses serviços, dependem da percepção e da experiência dos gestores responsáveis. Este trabalho visa tratar esse problema de estruturação das malhas de atendimento quanto à localização e alocação dos recursos chave (bases e viaturas) utilizando ferramentas e modelos matemáticos, de modo a propor uma formalização de todo o processo de planejamento. São revisados diversos modelos de localização de instalações e veículos de emergência e com base nas características dos grandes centros urbanos, propõe-se uma formulação matemática para o problema. Uma vez que esse modelo recai num problema de complexidade não polinomial, também é proposto um método de solução baseado no comportamento de colônias de abelhas na busca de alimentos. O método de soluções proposto é aplicado ao Serviço de Atendimento Móvel Pré-hospitalar de Urgência do Município de São Paulo em um estudo de caso. / Emergency systems are present in most of the large urban centers around the world. Its main goal is to provide assistance to victims of accidents, catastrophic events and general health issues in a quick and effective way. Therefore, a major characteristic of such systems is their resources readiness and a well-structured service network consisted basically by their emergency sites and ambulances. Operation of these systems is even more critical in large cities, in which the traffic conditions and the shifting pattern of demand along the day impose an extra complexity to the problem. Thus, an efficiently defined location/allocation map is crucial for the systems performance. In Brazil however, there is a gap between the processes, by which these maps are obtained and the facility location theories. Besides that, the methods currently applied to achieve a location/allocation map strongly depend on the perception and experience of decision makers. This work aims to deal with this planning problem by tackling the question with mathematical tools and models, resulting in a formal procedure for future emergency service network planning. A review of theories and mathematical models regarding emergency siting models is presented, and based on the large cities main characteristics, a model is proposed. Also, once this problem leads to a problem with non-polynomial complexity, a meta-heuristic was developed based on the behavior of foraging bees. This proposed solution method is also applied in a case study in São Paulo\'s Ambulance Service System (Serviço de Atendimento Móvel Pré-hospitalar de Urgência do Município de São Paulo).

Na maioria das cidades brasileiras, o ônibus é o principal, senão o único meio de transporte público de passageiros, atendendo àqueles que não possuem carro, mas também contribuindo para reduzir os congestionamentos e, dessa forma, melhorando a qualidade de vida. A fim de incentivar a sua utilização em muitas cidades, inúmeras medidas devem ser tomadas, incluindo um esforço para reduzir custos e, em decorrência, as tarifas pagas pelos usuários, definindo uma tarifa justa que privilegie a população de baixa renda, na qual estão inseridos uma grande parcela dos seus usuários. Nesse contexto, a programação eficiente de veículos e tripulações é essencial para essa redução de custos, uma vez que representa uma parcela significativa dos mesmos. Este trabalho trata do Problema de Programação de Veículos e de Tripulantes de Ônibus, em que ambas as programações são determinadas simultaneamente e de maneira integrada. Durante a realização deste trabalho, foi desenvolvido um programa computacional em C++, o qual implementa a metaheurística Busca em Vizinhança Variável (VNS) utilizada para resolver esse problema complexo. São apresentados resultados de experimentos computacionais com dados reais de uma empresa de transporte coletivo. Os resultados obtidos comprovam a eficácia da abordagem integrada em comparação com quatro outras abordagens encontradas na literatura. / In most Brazilian cities, bus has become the main, and oftentimes the only mean of public transportation, not only servicing those who cannot afford to have a car, but also helping to reduce traffic congestion and thus improving quality of life. In order to encourage the bus usage in many cities, several measures must be taken, including an effort to reduce overall costs and, as a result, fares paid by the users, thus benefit mainly low income users which represent a major part of its users. In this context, an efficient scheduling of vehicles and crews is essential to achieve cost reduction, since it accounts for a major part of overall costs. In this paper, we deal with the Bus Vehicle Crew Scheduling Problem, in which bus and crew schedules are simultaneously determined in an integrated approach. We propose an approach based on Variable Neighborhood search to solve this complex problem, which was implemented in C++. Computational results for real-world problems are presented, showing the effectiveness of this novel approach in comparison with other four approaches found in the literature.

O empacotamento irregular de fita é um grupo de problemas na área de corte e empacotamento, cuja aplicação é observada nas indústrias têxtil, moveleira e construção naval. O problema consiste em definir uma configuração de itens irregulares de modo que o comprimento do contêiner retangular que contém o leiaute seja minimizado. A solução deve ser válida, isto é, não deve haver sobreposição entre os itens, que não devem extrapolar as paredes do contêiner. Devido a aspectos práticos, são admitidas até quatro orientações para o item. O volume de material desperdiçado está diretamente relacionado à qualidade do leiaute obtido e, por este motivo, uma solução eficiente pressupõe uma vantagem econômica e resulta em um menor impacto ambiental. O objetivo deste trabalho consiste na geração automática de leiautes de modo a obter níveis de compactação e tempo de processamento compatíveis com outras soluções na literatura. A fim de atingir este objetivo, são realizadas duas propostas de solução. A primeira consiste no posicionamento sequencial dos itens de modo a maximizar a ocorrência de posições de encaixe, que estão relacionadas à restrição de movimento de um item no leiaute. Em linhas gerais, várias sequências de posicionamentos são exploradas com o objetivo de encontrar a solução mais compacta. Na segunda abordagem, que consiste na principal proposta deste trabalho, métodos rasterizados são aplicados para movimentar itens de acordo com uma grade de posicionamento, admitindo sobreposição. O método é baseado na estratégia de minimização de sobreposição, cujo objetivo é a eliminação da sobreposição em um contêiner fechado. Ambos os algoritmos foram testados utilizando o mesmo conjunto de problemas de referência da literatura. Foi verificado que a primeira estratégia não foi capaz de obter soluções satisfatórias, apesar de fornecer informações importantes sobre as propriedades das posições de encaixe. Por outro lado, a segunda abordagem obteve resultados competitivos. O desempenho do algoritmo também foi compatível com outras soluções, inclusive em casos nos quais o volume de dados era alto. Ademais, como trabalho futuro, o algoritmo pode ser estendido de modo a possibilitar a entrada de itens de geometria genérica, o que pode se tornar o grande diferencial da proposta. / Irregular nesting belongs to the area of cutting and packing problems and are employed in the textile, wood and shipbuilding industries. The problem consists in determining a configuration for a set of irregular items which minimizes the length of the rectangular container in which the layout is located. The solution must be feasible, i.e., items must not overlap nor protrude the container walls. Due to practical reasons, up to four orientations are allowed for an item. The volume of wasted material is directly affected by the quality (density) of the layout. Thus, an efficient solution produces a positive economic and environmental impact. In this work, the objective is to automatically obtain layouts such that their density and the performance of the algorithm are competitive with other solutions in literature. So as to achieve this goal, two approaches are proposed. The first method uses a special sequential placement heuristic such that the algorithm maximizes exact placements, which consist of constrained positions for items. In general terms, a search is performed in the placement sequence in order to obtain a compact layout. In the second approach, which is the main subject of this work, raster methods are employed to guide the translation of items, which are free to move within the layout, and may overlap other items. The method is based on overlap minimization techniques, in which the objective is to eliminate the overlap in a fixed dimensions container. Both algorithms were tested using benchmark problems from the literature. The first strategy yielded unsatisfactory results, though it provided important information about the properties of exactly fitting placements. On the other hand, the main approach was able to produce competitive solutions. The performance was also compatible with other solutions, even in cases which the data volume was high. Moreover, as a future work, an extension for the algorithm can be developed such that items with generic geometry can be considered, which would be an important advance in research terms.