Heterogeneous Computing and Parallel Genetic Algorithms (original) (raw)

A Heterogeneous Framework for the Global Parallelisation of Genetic Algorithms

There is a big need for the parallelisation of genetic algorithms. In this paper, a heterogeneous framework for the global parallelisation of genetic algorithms is presented. The framework uses a static all-worker parallel programming paradigm based on collective communication. It follows the single program multiple data parallel programming model. It utilises the power of parallel machines by allowing multiple crossover and mutation operators being used within a single genetic algorithm. This mixture of operators can be applied to the strings of a population in parallel without changes to the canonical sequential genetic algorithm. These features help the parallel genetic algorithm in exploiting the search space efficiently and thoroughly when compared to the sequential genetic algorithm. The framework is instantiated with specific parameters to solve an NP-hard problem, the asymmetric travelling salesman problem. The results for the parallel genetic algorithm are very good in term...

Creation and Analysis of a JavaSpace-based Distributed Genetic Algorithm

2002

The island model for distributed genetic algorithms (GAs) is a natural match for the master-worker paradigm in distributed computation. We explore the benefits and drawbacks of several distributed system architectures in developing an implementation of a distributed GA that exploits the Jini and JavaSpace technologies. Our results, using the knapsack problem as an illustration, show that there is an unavoidable price to pay in terms of decreasing computation-to-communication ratios as a function of instance size. However, we can diminish these effects by expanding the number of JavaSpaces beyond those required for the obvious implementation. Our results also indicate that as the number of remote machines increases the potential for a better solution also rises. Even though our distributed GAs did not always exploit this potential for a higher quality solution, we believe that the combination of Java, Jini, and JavaSpaces presents avenues for easily distributing the computation of genetic algorithms.

Parallel Genetic Algorithms In A NetworkedWorkstation Environment

WIT Transactions on Information and Communication Technologies, 1997

Parallel Genetic Algorithms are suited to deal with problems with very large solution spaces and they can support efficient parallel distribution of work. In a PGA Island Model the migration strategy can take advantage of high latency communication channels in a distributed system. This approach suggests the use of networked workstation environments as a cost effective alternative to MPP systems. A Genetic Algorithm Programming System (GAPS) was developed to evaluate the proposed approach, which supports the design of parallel genetic programs and its execution in a distributed workstation environment. GAPS separates the specification of the problem and the user application interface, from the implementation and management details of the run-time environment; it also addresses fault tolerance, needed to recover from a fault that may occur in a dynamic network of heterogeneous workstations. GAPS uses PVM to implement a structural load balance strategy, which distributes complex evalu...

Scalable Parallel Genetic Algorithms

Artificial Intelligence Review, 2001

Genetic algorithms, search algorithms based on the genetic processes observed in natural evolution, have been used to solve difficult problems in many different disciplines. When applied to very large-scale problems, genetic algorithms exhibit high computational cost and degradation of the quality of the solutions because of the increased complexity. One of the most relevant research trends in genetic algorithms is

A Java-Based Distributed Genetic Algorithm Framework

Tools with Artificial …, 2007

Distributed Genetic Algorithm (DGA) is one of the most promising choices among the optimization methods. In this paper we describe DGAFrame, a flexible framework for evolutionary computation, written in Java. DGAFrame executes GAs across a range of machines communicating through RMI network technology, allowing the implementation of portable, flexible GAs that use the island model approach. Each island can be configured independently from others providing the implementation of heterogeneous DGAs. To evaluate the performance of DGAFrame, we implemented the Protein Structure Prediction problem and compare the DGA execution to its sequential counterpart through quality of solution. We also measure the computation to communication ratio and results show that the proposals consistently outperform equivalent sequential GAs.

Parallel genetic algorithms on line topology of heterogeneous computing resources

… of the 2005 conference on Genetic …, 2005

This paper evaluates a parallel genetic algorithm (GA) on the line topology of heterogeneous computing resources. Evolution process of parallel GAs is investigated on two types of arrangements of heterogeneous computing resources: the ascending and descending order arrangement of computing capability. Their differences in chromosome variety, migration frequency and solution quality are investigated. The results in this paper can help to design parallel GAs in grid computing environments.

Efficient Parallel Genetic Algorithms: Theory and Practice

2000

Parallel genetic algorithms (GAs) are complex programs that are controlled by many parameters, which a ect their search quality and their e ciency. The goal of this paper is to provide guidelines to choose those parameters rationally. The investigation centers on the sizing of populations, because previous studies show that there is a crucial relation between solution quality and population size. As a rst step, the paper shows how to size a simple GA to reach a solution of a desired quality. The simple GA is then parallelized, and its execution time is optimized. The rest of the paper deals with parallel GAs with multiple populations. Two bounding cases of the migration rate and topology are analyzed, and the case that yields good speedups is optimized. Later, the models are specialized to consider sparse topologies and migration rates that are more likely to be used by practitioners. The paper also presents the additional advantages of combining multi-and single-population parallel GAs. The results of this work are simple models that practitioners may use to design e cient and competent parallel GAs.

A Survey: Genetic Algorithms and the Fast Evolving World of Parallel Computing

2008 10th IEEE International Conference on High Performance Computing and Communications, 2008

. Helping the GA community to feel more comfortable with the evolving parallel paradigms, and marking some areas of research for the High-Performance Computing (HPC) community is the major inspiration behind this survey. In the modern parallel computing paradigms we have considered only two major areas that have evolved very quickly during the past few years, namely, multicore computing and Grid computing. We discuss the challenges involved, and give potential solutions for these challenges. We also propose a hierarchical PGA suitable for Grid environment with multicore computational resources.

A Survey of Parallel Genetic Algorithms

1997

Genetic algorithms (GAs) are powerful search techniques that are used successfully to solve problems in many different disciplines. Parallel GAs are particularly easy to implement and promise substantial gains in performance. As such, there has been extensive research in this field. This survey attempts to collect, organize, and present in a unified way some of the most representative publications on parallel genetic algorithms. To organize the literature, the paper presents a categorization of the techniques used to parallelize GAs, and shows examples of all of them. However, since the majority of the research in this field has concentrated on parallel GAs with multiple populations, the survey focuses on this type of algorithms. Also, the paper describes some of the most significant problems in modeling and designing multi-population parallel GAs and presents some recent advancements.