No abstract

The effects of various population topologies on the particle swarm algorithm were systematically investigated. Random graphs were generated to specifications, and their performance on several criteria was compared. What makes a good population structure? We discovered that previous assumptions may not have been correct.

This paper presents an approach of using Differential Evolution (DE) to solve dynamic optimization problems. Careful setting of parameters is necessary for DE algorithms to successfully solve optimization problems. This paper describes DynDE, a multi-population DE algorithm developed specifically to solve dynamic optimization problems that doesn't need any parameter control strategy for the F or CR parameters. Experimental evidence has been gathered to show that this new algorithm is capable of efficiently solving the moving peaks benchmark.

Background: One of the greatest challenges in Metabolic Engineering is to develop quantitative models and algorithms to identify a set of genetic manipulations that will result in a microbial strain with a desirable metabolic phenotype which typically means having a high yield/productivity. This challenge is not only due to the inherent complexity of the metabolic and regulatory networks, but also to the lack of appropriate modelling and optimization tools. To this end, Evolutionary Algorithms (EAs) have been proposed for in silico metabolic engineering, for example, to identify sets of gene deletions towards maximization of a desired physiological objective function. In this approach, each mutant strain is evaluated by resorting to the simulation of its phenotype using the Flux-Balance Analysis (FBA) approach, together with the premise that microorganisms have maximized their growth along natural evolution.

We vary the way an individual in the particle swarm interacts with its neighbors. Performance depends on population topolog as well as algorithm version.The particle swarm algorithm is based on a socialpsychological model of social influence and social learning. A population of candidate problem solutions, randomly initialized in a high-dimensional search space, discovers optimal regions of the space through a process of individuals' emulation of the successes of their neighbors. The present paper investigates an alternative implementation of social neighborhoods within the particle swarm framework.In the traditional particle swarm, each individual has some number of neighbors, with mutual influence between them. On each iteration of the program loop, the individual queries its neighbors to determine which one has had the best success with the problem thus far, and uses the location of that success, plus the location of its own previous best success, to choose a new point in the search space to test.This represents an oversimplification of the socialpsychological view that individuals are more affected by sources of influence who are most successful, persuasive, or otherwise prestigious. In human society it is more accurate to say that the social neighborhood provides a wealth of possible models whose behavior may be emulated, and individuals seem to be affected by some kind of statistical summary of the state of their immediate social network rather than the unique performance of one individual (e.g., Latane, 1981; Granovetter, 1977; etc.). Thus, the present paper reports on research with a version of particle swarm where the individual is influenced by the successes of all its neighbors, rather than just the best one. The fullyinformed particle swarm (FIPS) is compared to the canonical version, using a variety of neighborhood structures. NeighborhoodsThe earliest reported particle swarm version Eberhart and Kennedy, 1995) used a kind of topology that became known as gbest.The source of social influence on each particle was the Rui MendesUniversidade do Minho Braga, Portugal azuki@di.uminho.pt best-performing individual in the entire population. This is equivalent to a sociogram or social network where every individual is connected to every other one.The gbest topology was acceptable for the first applications, which typically involved finding a matrix of weights for a feedforward neural network. The hnction landscape of this kind of problem is largely made up of long gradients, where the problem is, first, to find the best gradient region of the search space, and second, to find the minimum of that region.But many problems contain cliffs, variable interactions, and other features that are not typified by smooth gradients, and a more robust algorithm was needed. The lbest topology was proposed as a way to deal with more difficult problems. In the lbest sociometric structure, each individual is connected to -that is, influences and is influenced by -its immediate neighbors in the population array. The fifth ind...