Font Size: a A A

Research On The Bacterial Foraging Optimization Algorithm

Posted on:2013-11-20Degree:DoctorType:Dissertation
Country:ChinaCandidate:X XuFull Text:PDF
GTID:1228330395459636Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Bacterial foraging optimization algorithm (BFO) is a meta-heuristic algorithm, which isinspired by the chemotactic behavior of bacterial such as E.coli and M.xanthus. BFO iscurrently getting more and more popular in the community of researchers, for its effectivenessin solving certain difficult real-world optimization problems. It is pointed that BFO has theglobal search capability for dealing with the low-dimensional continuous optimizationproblems. However, as can be shown from the literature, BFO was rarely used to solve thecombinatorial optimization problems (COPs). At the same time, the optimization performanceof BFO is reduced as the dimension and complexity of the optimization problems areincreased. Thus, how to identify the key factors which affect the capbality of BFO, andproviding some robust and efficient improved versions of BFO for high-dimensionaloptimization problems and COPs, which have not yet been well solved up to now.This paper studies the construction of the improved BFO with strong generalizationability. We focus on studying the four key factors such as adaptive chemotactic step size,swarm operator, multi-objective BFO and discrete BFO that affect the capability of BFO. Wehave done the systematic and in-depth study on inherent defects of the improved BFO whichinvolves these factors. We also carried out a series of studies by focusing on how to buildseveral novel improved versions of BFO with robust and efficiency. We have proposed theadapive computational chemotaxis in bacterial foraging optimization algorithm base on thefield, the adaptive bacterial swarm optimization algorithm with time-varying accelerationcoefficients, the multi-objective bacterial swarm optimization algorithm, and the set-basedadaptive bacterial swarm optimization algorithm. The main contribution of this paper andinnovative points are as follows:(1) We have done a comprehensive overview on the state-of-the-art of the BFO, made ananalysis and discussion of problems faced and the current development trends in this area. Inaddition, we have made a brief overview of three famous meta-heuristic algorithms such asgenetic algorithm (GA), particle swarm optimization (PSO), and ant colony optimization(ACO). We also have compared the difference between the BFO and each of them. Thediscussion and analysis of these topics have laid a solid foundation for carring out furtherresearch work.(2) This paper designs an improved version of adaptive BFO. One of the majorcharacteristics of BFO is the chemotactic movement of a virtual bacterium that models a trialsolution of the problems. It is pointed out that the chemotaxis employed by classical BFOusually results in sustained oscillation, especially on rough fitness landscapes, when abacterium cell is close to the optima. In this paper we propose a novel adaptive computational chemotaxis based on the concept of field, in order to accelerate the convergence speed of thegroup of bacteria near the tolerance. Firstly, a simple scheme is designed for adapting thechemotactic step size of each field. Then, the scheme chooses the fields which perform betterto boost further the convergence speed. Empirical simulations over several numericalbenchmarks demonstrate that BFO with adaptive chemotactic operators based on field hasbetter convergence behavior, as compared against other meta-heuristic algorithms.(3) This paper also provides an adaptive bacterial swarm optimization algorithm forhigh-dimensional problems. Until now, there is much room for improvement in convergencespeed and accuracy, because the distances of all the bacteria in the chemotactic stage are notused to select the direction for the bacteria to run. In addition, several hybrid approachesintegrating BFO with other meta-heuristics methods have been proposed to improve theconvergence speed and accuracy of basic BFO. However, the idea behind these hybridschemes lies in implementing each meta-heuristic algorithm in turn one by one, thus thepotential of the BFO can not be fully explored.In this paper, we propose a novel adaptive bacterial swarm optimization algorithm withtime-varying acceleration coefficients, termed as ABSO-TVAC, in order to further acceleratethe convergence speed and accuracy of the adaptive BFO. Firstly, a novel swarming operationis designed for searching the optimal solutions on each field which is comprised of two orthree dimensional space. Then, a novel chemotaxis mechanism, which is inspired by theconcept of hierarchical particle swarm optimization algorithm with time-varying accelerationcoefficients, is proposed for controlling the global search and converging to the globaloptimum. Empirical simulations over several numerical benchmarks demonstrate theproposed ABFO-TVAC has shown much better convergence behavior, as compared againstother meta-heuristic algorithms.(4) This paper proposes an adaptive multi-objective bacterial swarm optimizationalgorithm to solve multi-objective problems. The existing multi-objective BFO (MBFO) canfind a little non-inferior solution of the Pareto front, becasuce the MBFO does not completelytransform all operators of the original BFO. To make up for this shortcoming, this paperproposes an adaptive multi-objective bacterial swarm optimization algorithm (AMBSO) formulti-objective problems. The proposed AMBSO method implements the search for Paretooptimal set of multi-objective optimization problems. The AMBSO has been compared withthe MBFO over a test suite of five ZDT numerical benchmarks with respect to the twoperformance measures: Generational Distance and Diversity Measure. The simulation resultsshow that the AMBSO is able to find a much better Pareto front solutions.(5) This paper provides a set-based adaptive bacterial swarm optimization algorithm(S-BSO) for the COPs. Firstly, according to the cell-to-cell signaling in E.coli swarm, wedesigned a conversion rule between the continuous space and discrete one. Then, allarithmetic operators in the velocity and position updating rules used in the ABSO-TVAC areall descried in a manner of set. Simulation shows that the proposed algorithm has the capability to avoid the premature problem, and the result is superior to ACO and comparableto set-based PSO.In this paper, we have designed an improved BFO, and proposed three versions ofbacterial swarm algorithms to solve high-dimensional continuous optimization problems,multi-objective continous optimization problems, and COPs, respectively. These studies notonly promote the futher development of the BFO, and also laid the solid foundation for theapplication of the BFO in the area of decision support. Therefore, this study has importanttheoretical significance and application value as well.
Keywords/Search Tags:Artificial intelligence, Bacterial foraging optimization algorithm, Adaptivechemotaxis size, Swarm operator, Multi-objective bacterial foraging optimization algorithm, Discrete bacterial foraging optimization algorithm
PDF Full Text Request
Related items