Parallel Island Model Research Articles

On reconfiguring heterogeneous parallel island models

This work introduces a new Parallel Island Model (PIM) that encompasses the benefits of heterogeneity and algorithmic reconfigurability. The former feature, heterogeneity, means that different islands may execute different evolutionary algorithms. The latter, reconfigurability, means that each island can change the algorithm being executed during the evolutionary process. Allowing such features increases the usual diversity obtained by the communication topologies and migration policies by homogeneous PIMs (HoPIMs). Previous (non-reconfigurable) heterogeneous PIMs (HePIMs) were able to provide competitive solutions regarding the HoPIMs. By adding the reconfiguration capability, PIMs can change dynamically from executing one evolutionary algorithm to another. In this manner, the required diversity and flexibility to outperform HoPIMs and HePIMs is achieved. This paper discusses policies to profit from the feature of reconfigurability on HePIM models and provides an innovative and successful stagnation-based reconfiguration policy. The benefits of the new reconfigurable model are verified using the unsigned reversal distance optimization problem as a case study.

On the behavior of parallel island models

Parallel island models (PIMs) are used to enhance the performance of evolutionary algorithms (EAs). In PIMs, each island executes an EA for evolving its local population, and periodically individuals migrate to neighborhoods synchronously or asynchronously. Neighborhoods are organized through a topology of communication, and migration policies guide individual exchange. This work explores migration policies over different communication topologies in synchronous and asynchronous PIMs to improve the speed-up and accuracy of genetic algorithms (GAs). The aim is to explain such models’ adequacy from a general perspective, attempting to answer questions such as the best way to implement GAs in PIMs. To reach this goal, the quality of the solutions and the running time provided by proposed PIMs are evaluated over four NP-hard problems of different combinatorial nature: reversal and translocation evolutionary distance, task mapping and scheduling, and N-Queens. The experiments show that tuning the parameters of the breeding cycle and migration policies is vital to guarantee that all proposed PIMs reach good speed-ups and more accurate solutions than the sequential GA. In addition, experiments ratify that synchronous models provide the best solutions, while asynchronous models deliver the best speed-ups. Finally, the results show that no model provides either better speed-up or accuracy in general since, as for sequential models, the nature of the problem define which would be the best-adapted PIM.

Optimal Sensors/Actuators Placement in Smart Structure Using Island Model Parallel Genetic Algorithm

Determination of optimal placements of sensors/actuators in large structures is a difficult job as large number of possible combinations leads to a very high computational time and storage. Therefore, this kind of optimization problem demands a parallel implementation of the optimization schemes. Island model genetic algorithm (GA) being inherently parallel has been used for searching optimal placements of collocated sensors/actuators. Numerical simulations have been done for determination of optimal placements of collocated PZT sensors and actuators in smart fiber reinforced shell structures using island model parallel GA (IMPGA) in conjunction with electro-mechanical finite element analysis with an objective of maximizing the controllability index. It has been observed that the present IMPGA-based formulation (due to its migration scheme) not only makes it possible to determine optimal sensors/actuators locations for large structures but also leads to a better solution at a much reduced and achievable computational time. Results from scalability analysis also show that the efficacy of the present method of using IMPGA for determination of optimal sensors/actuators location based on FEA will be more pronounced when actually used for real life problems requiring large number of sensors and actuators.

An island strategy for memetic discrete tomography reconstruction

In this paper we present a parallel island model memetic algorithm for binary discrete tomography reconstruction that uses only four projections without any further a priori information. The underlying combination strategy consists in separated populations of agents that evolve by means of different processes. Agents progress towards a possible solution by using genetic operators, switch and a particular compactness operator. A guided migration scheme is applied to select suitable migrants by considering both their own and their sub-population fitness. That is, from time to time, we allow some individuals to transfer to different subpopulations. The benefits of this paradigm were tested in terms of correctness, robustness and time of the reconstruction by considering publicly available datasets of images. To tackle the so-called stability problem, we considered the case of noisy projections along four directions to simulate an instrumental error. Results show that the proposed method decreases the reconstruction error for all classes of images with respect to a serial implementation recently proposed by the authors, and that such reconstruction error is almost invariant with respect to the number of demes. Moreover, the computation time of the proposed parallel memetic algorithm scales in a quasi-linear manner with respect to the demes number, and is invariant with respect to the used number of migrations.

Contribution of parallel NSGA-II in optimal design of water distribution networks

Optimization of water distribution networks is a NP-hard problem that researchers have tried to deal with using different formulations and algorithmic approaches. Among these, multi-objective heuristic algorithms are interesting because of their capacity for dealing with separate objectives that allow us to choose a posteriori the best compromise, but one of their main drawbacks is the long time required to obtain good solutions. Parallel processing is the most promising way to reduce the computing time and can make the convergence to adequate solutions faster. This paper intends to investigate the possibility of improving the efficacy and efficiency of an NSGA-II algorithm by parallelization of the optimization process at the same time. Results of different parallel implementations of NSGA-II applied to optimal design of small- and medium-size water distribution networks are presented. Good speed-up can be reached with a global model, hence improving the algorithm efficiency. Unlike the global model, the island model (or the hierarchical parallelization) can also improve its efficacy because it introduces fundamental changes in the algorithm exploration method. Possibilities offered by parallel island models have been investigated showing that some parameter configurations can find better solutions compared with the serial version of the algorithm.

Synthesis of Time-to-Amplitude Converter by Mean CoeVolution with Adaptive Parameters

The challenging task to synthesize automatically a time-to-amplitude converter, which unites by its functionality several digital circuits, has been successfully solved with the help of a novel methodology. The proposed approach is based on a paradigm according to which the substructures are regarded as additional mutation types and when ranged with other mutations form a new adaptive individual-level mutation technique. This mutation approach led to the discovery of an original coevolution strategy that is characterized by very low selection rates. Parallel island-model evolution has been running in a hybrid competitive-cooperative interaction throughout two incremental stages. The adaptive population size is applied for synchronization of the parallel evolutions.

An overview of evolutionary algorithms: practical issues and common pitfalls

An overview of evolutionary algorithms is presented covering genetic algorithms, evolution strategies, genetic programming and evolutionary programming. The schema theorem is reviewed and critiqued. Gray codes, bit representations and real-valued representations are discussed for parameter optimization problems. Parallel Island models are also reviewed, and the evaluation of evolutionary algorithms is discussed.

A genetic algorithm tutorial

This tutorial covers the canonical genetic algorithm as well as more experimental forms of genetic algorithms, including parallel island models and parallel cellular genetic algorithms. The tutorial also illustrates genetic search by hyperplane sampling. The theoretical foundations of genetic algorithms are reviewed, include the schema theorem as well as recently developed exact models of the canonical genetic algorithm.

The Backboard Wiring Problem: A Placement Algorithm

The Backboard Wiring Problem: A Placement Algorithm