Made available in DSpace on 2016-03-15T19:38:16Z (GMT). No. of bitstreams: 1 Marcio Frayze David.pdf: 699315 bytes, checksum: 587538708d29252e3c3a8f5c46cbaa53 (MD5) Previous issue date: 2010-05-03 / Fundo Mackenzie de Pesquisa / This dissertation discusses the use of bio-inspired algorithms for data clustering, with emphasis on a model of emergent collective behavior of agents and a new clustering algorithm called cBoids is presented. The cBoids algorithm is a variation of the classic Boids model. In this new algorithm, each Boid represents an object from the data base and the three original rules from the Boids model were modified so that the objects of the database have influence on the behaviour of the Boids. Two new rules have also been proposed, responsible for the creation and destruction of centroids, which represent the formed clusters. In the experiments conducted in this work the algorithm was successfully tested on four databases. / Esta dissertação aborda o uso de algoritmos bioinspirados para a tarefa de agrupamento de dados , com ênfase nos modelos de comportamentos emergentes coletivos de agentes e um novo algoritmo de agrupamento de dados chamado cBoids é apresentado. O algoritmo cBoids é uma variação do clássico modelo Boids. Neste novo algoritmo, cada Boid representa um objeto da base de dados e as três regras originais do modelo Boids foram alteradas para que os objetos da base de dados influenciem o comportamento dos Boids. Duas novas regras também foram propostas, responsáveis pela criação e destruição de centróides, que representam os clusters formados. Nos experimentos realizados nesta dissertação o algoritmo foi testado com sucesso em quatro bases de dados.

Context. Agent based models are used to simulate complex systems by using multiple agents that follow a set of rules. One such model is the boid model which is used to simulate movements of synchronized groups of animals. Executing agent based models partially or fully on the GPU has previously shown to increase performance, opening up the possibility for larger simulations. However, few articles have previously compared a full GPU implementation of the boid model with a multi-threaded CPU implementation. Objectives. The objectives of this thesis are to find how parallel execution of boid model performs when executed on the CPU and GPU respectively, based on the variables frames per second and average boid computation time per frame. Methods. A performance benchmark experiment will be set up where three implementations of the boid model are implemented and tested. Results. The collected data is summarized in both tables and graphs, showing the result of the experiment for frames per second and average boid computation time per frame. Additionally, the average results are summarized in two tables. Conclusions. For the largest flock size the GPGPU implementation performs the best with an average FPS of 42 times over the single-core implementation while the multi-core implementation performs with an average FPS 6 times better than the single-core implementation. For the smallest flock size the single-core implementation is most efficient while the GPGPU implementation has 1.6 times slower average update time and the multi-cor eimplementation has an average update time of 11 times slower compared to the single-core implementation.

Boid-based systems are typically used to simulate flocks of animals in nature, such as avian flocks or fish schools. The naïve implementation has a time complexity of $O(n^2)$, as each boid perceives and reacts to every other, and thus scales poorly with flock size. Several optimizations and strategies have previously been explored to mitigate this fact, such as space partitioning of the game world where the simulation runs. This thesis explores the performance bottlenecks in an implementation utilizing some of these optimizations, mainly spatial hashing. The findings indicate that the major performance bottleneck is the aggregation of positions and headings required to enact the boid steering behaviors. This performance bottleneck can be alleviated by spatially hashing the game world, using a cell-based neighborhood, and parallelizing the computations over these cells. In addition, parameters controlling the simulation can be tweaked and upper-bound limits imposed on the number of interactions to further improve performance.