Encontre seu próximo livro favorito

Torne'se membro hoje e leia gratuitamente por 30 dias.
GPU-based Parallel Implementation of Swarm Intelligence Algorithms

GPU-based Parallel Implementation of Swarm Intelligence Algorithms

Ler amostra

GPU-based Parallel Implementation of Swarm Intelligence Algorithms

434 página
Lançado em:
Apr 15, 2016


GPU-based Parallel Implementation of Swarm Intelligence Algorithms combines and covers two emerging areas attracting increased attention and applications: graphics processing units (GPUs) for general-purpose computing (GPGPU) and swarm intelligence. This book not only presents GPGPU in adequate detail, but also includes guidance on the appropriate implementation of swarm intelligence algorithms on the GPU platform.

GPU-based implementations of several typical swarm intelligence algorithms such as PSO, FWA, GA, DE, and ACO are presented and having described the implementation details including parallel models, implementation considerations as well as performance metrics are discussed. Finally, several typical applications of GPU-based swarm intelligence algorithms are presented. This valuable reference book provides a unique perspective not possible by studying either GPGPU or swarm intelligence alone.

This book gives a complete and whole picture for interested readers and new comers who will find many implementation algorithms in the book suitable for immediate use in their projects. Additionally, some algorithms can also be used as a starting point for further research.

  • Presents a concise but sufficient introduction to general-purpose GPU computing which can help the layman become familiar with this emerging computing technique
  • Describes implementation details, such as parallel models and performance metrics, so readers can easily utilize the techniques to accelerate their algorithmic programs
  • Appeals to readers from the domain of high performance computing (HPC) who will find the relatively young research domain of swarm intelligence very interesting
  • Includes many real-world applications, which can be of great help in deciding whether or not swarm intelligence algorithms or GPGPU is appropriate for the task at hand
Lançado em:
Apr 15, 2016

Sobre o autor

Ying Tan is a professor and PhD advisor at the School of Electronics Engineering and Computer Science of Peking University, and director of Computational Intelligence Laboratory at Peking University (PKU). He received his BEng from the EEI, MSc from Xidian Univ., and PhD from Southeast Univ., in 1985, 1988, and 1997, respectively. From 1997, he was a postdoctoral fellow then an associate professor at University of Science and Technology of China (USTC), then served as director of Institute of Intelligent Information Science and a full professor since 2000. He worked with the Chinese University of Hong Kong (CUHK) in 1999 and 2004-2005. He was elected for the 100 talent program of the Chinese Academy of Science (CAS) in 2005

Relacionado a GPU-based Parallel Implementation of Swarm Intelligence Algorithms

Livros relacionados
Artigos relacionados

Amostra do Livro

GPU-based Parallel Implementation of Swarm Intelligence Algorithms - Ying Tan


Chapter 1



1.1 Swarm Intelligence Algorithms (SIAs)  1

1.2 Graphics Processing Units (GPUs)  3

1.3 SIAs and GPUs  3

1.4 Some Perspectives  5

1.5 Organization  6

1.1 Swarm Intelligence Algorithms (SIAs)

Swarm intelligence is the collective behavior of decentralized, self-organized systems. A typical swarm intelligence system consists of a population of simple agents which can communicate (either directly or indirectly) locally with each other by acting on their local environment. Though the agents in a swarm follow very simple rules, the interactions between such agents can lead to the emergence of very complicated global behavior, far beyond the capability of individual agents [54, 55]. Examples in natural systems of swarm intelligence include bird flocking, ant foraging, and fish schooling.

Inspired by swarm’s such behavior, a class of algorithms is proposed for tackling optimization problems, usually under the title of swarm intelligence algorithms (SIAs) [203]. In SIAs, a swarm is made up of multiple artificial agents. The agents can exchange heuristic information in the form of local interaction. Such interaction, in addition with certain stochastic elements, generates the behavior of adaptive search, and finally leads to global optimization.

The most respected and popular SIAs are particle swarm optimization (PSO) which is inspired by the social behavior of bird flocking or fish schooling [87, 19], and ant colony optimization (ACO) which simulates the foraging behavior of ant colony [52, 51]. PSO is widely used for real-parameter optimization while ACO has been successfully applied to solve combinatorial optimization problems, for instance, the most well-known combinatorial optimization problems are the traveling salesman problem (TSP) and quadratic assignment problem (QAP).

Novel SIAs with particular search mechanisms have been proposed and achieved success on specific problems. Some instances of novel SIAs are bacterial foraging optimization (BFO) [145], bee algorithm [148], fish schooling search (FSS) [57], cuckoo search [204], fireworks algorithm (FWA) [180–182, 185], brain storm optimization (BSO) [174], and the list is increasingly long.

In essence, SIAs are iterative-based stochastic search algorithms where heuristic information is shared in order to guide the search in the following iterations. A simplified general framework of SIAs is depicted in Fig. 1.1. For a particular SIAs, the sequence of each phase may be different and some phases can be included several times in a single iteration.

Fig. 1.1 Framework of Swarm Intelligence Algorithms

Though different SIAs utilize different interaction mechanisms and various heuristic information in quite different ways, we think this framework is capable of embodying the essential philosophy underlying SIAs, thus can be used as the reference frame for discussing various parallel implementations on the graphics processing unit (GPU).

Take standard PSO for example. After randomly initializing the population, the fitness value of each particle is evaluated. In the following communication phase, fitness values are communicated among particles according to certain topology (global topology, ring topology, etc.). Then each particle updates its private and neighborhood best information (if necessary), and its velocity and position vector in swarm updating phase. The loop continues until termination conditions (e.g., meeting the presetting accuracy or reaching the maximum number of fitness evaluations) are satisfied.

In the case of typical ACO, ants communicate with one another indirectly via pheromone updating. Each ant needs to construct a feasible path before it can evaluate the fitness (cost of the path). This path construction step may be quite different from PSO and other SIAs for real-parameter optimization at the first glance. However, thinking this step under the proposed reference frame, the path construction is nothing but a stochastic updating mechanism. As the ant has no private memory, it must build the new solution from scratch with the pheromone as heuristic guide. Once the path is constructed, the calculation of the fitness value is trivial. More often, for combinatorial optimization problems, the fitness value is calculated as the same time as the path is under construction. In this case, fitness evaluation phase and swarm updating can be viewed as emerged into one single

Você chegou ao final desta amostra. Inscreva-se para ler mais!
Página 1 de 1


O que as pessoas pensam sobre GPU-based Parallel Implementation of Swarm Intelligence Algorithms

0 avaliações / 0 Análises
O que você acha?
Classificação: 0 de 5 estrelas

Avaliações de leitores