Particle Swarm Optimization with Adaptive Mutation in Local Best of Particles

Similar documents
The Study of Teaching-learning-based Optimization Algorithm

An Adaptive Learning Particle Swarm Optimizer for Function Optimization

A New Quantum Behaved Particle Swarm Optimization

Riccardo Poli, James Kennedy, Tim Blackwell: Particle swarm optimization. Swarm Intelligence 1(1): (2007)

MODIFIED PARTICLE SWARM OPTIMIZATION FOR OPTIMIZATION PROBLEMS

An improved multi-objective evolutionary algorithm based on point of reference

Comparative Analysis of SPSO and PSO to Optimal Power Flow Solutions

Entropy Generation Minimization of Pin Fin Heat Sinks by Means of Metaheuristic Methods

V is the velocity of the i th

Research Article Multiobjective Economic Load Dispatch Problem Solved by New PSO

Evolutionary Computational Techniques to Solve Economic Load Dispatch Problem Considering Generator Operating Constraints

Research Article Adaptive Parameters for a Modified Comprehensive Learning Particle Swarm Optimizer

A New Evolutionary Computation Based Approach for Learning Bayesian Network

A Hybrid Co-evolutionary Particle Swarm Optimization Algorithm for Solving Constrained Engineering Design Problems

Using Immune Genetic Algorithm to Optimize BP Neural Network and Its Application Peng-fei LIU1,Qun-tai SHEN1 and Jun ZHI2,*

Open Access A Variable Neighborhood Particle Swarm Algorithm Based on the Visual of Artificial Fish

Exploratory Toolkit for Evolutionary And Swarm based Optimization

A PROBABILITY-DRIVEN SEARCH ALGORITHM FOR SOLVING MULTI-OBJECTIVE OPTIMIZATION PROBLEMS

Optimal Placement of Unified Power Flow Controllers : An Approach to Maximize the Loadability of Transmission Lines

Design and Optimization of Fuzzy Controller for Inverse Pendulum System Using Genetic Algorithm

PARTICLE SWARM OPTIMIZATION BASED OPTIMAL POWER FLOW FOR VOLT-VAR CONTROL

Differential Evolution Algorithm with a Modified Archiving-based Adaptive Tradeoff Model for Optimal Power Flow

Coke Ratio Prediction Based on Immune Particle Swarm Neural Networks

Markov Chain Monte Carlo Lecture 6

A Novel Hybrid Bat Algorithm for the Multilevel Thresholding Medical Image Segmentation

A novel hybrid algorithm with marriage of particle swarm optimization and extremal optimization

A Study on Improved Cockroach Swarm Optimization Algorithm

A Self-Adaptive Chaos Particle Swarm Optimization Algorithm

Appendix B: Resampling Algorithms

Solving of Single-objective Problems based on a Modified Multiple-crossover Genetic Algorithm: Test Function Study

An Improved Particle Swarm Optimization Algorithm based on Membrane Structure

Self-adaptive Differential Evolution Algorithm for Constrained Real-Parameter Optimization

Chapter - 2. Distribution System Power Flow Analysis

A Hybrid Differential Evolution Algorithm Game Theory for the Berth Allocation Problem

A Particle Swarm Algorithm for Optimization of Complex System Reliability

Multi-objective Optimization of Condensation Heat Transfer inside Horizontal Tube using Particle Swarm Optimization Technique

A NEW METHOD TO INCORPORATE FACTS DEVICES IN OPTIMAL POWER FLOW USING PARTICLE SWARM OPTIMIZATION

Research on Route guidance of logistic scheduling problem under fuzzy time window

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Lecture Notes on Linear Regression

Wavelet chaotic neural networks and their application to continuous function optimization

An adaptive SMC scheme for ABC. Bayesian Computation (ABC)

Optimal Reactive Power Dispatch Using Efficient Particle Swarm Optimization Algorithm

APPLICATION OF RBF NEURAL NETWORK IMPROVED BY PSO ALGORITHM IN FAULT DIAGNOSIS

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

A Hybrid Variational Iteration Method for Blasius Equation

Comparison of the Population Variance Estimators. of 2-Parameter Exponential Distribution Based on. Multiple Criteria Decision Making Method

An Improved Clustering Based Genetic Algorithm for Solving Complex NP Problems

INTERNATIONAL JOURNAL OF ELECTRICAL ENGINEERING & TECHNOLOGY (IJEET)

Particle Swarm Optimization for Non-Convex Problems of Size and Shape Optimization of Trusses

arxiv: v1 [math.oc] 3 Aug 2010

Transient Stability Constrained Optimal Power Flow Using Improved Particle Swarm Optimization

A New Scrambling Evaluation Scheme based on Spatial Distribution Entropy and Centroid Difference of Bit-plane

(Online First)A Lattice Boltzmann Scheme for Diffusion Equation in Spherical Coordinate

Solving Design of Pressure Vessel Engineering Problem Using a Fruit Fly Optimization Algorithm

OPTIMAL POWER FLOW USING PARTICLE SWARM OPTIMIZATION

GHHAGA for Environmental Systems Optimization

Implementation of digital pheromones for use in particle swarm optimization

FUZZY GOAL PROGRAMMING VS ORDINARY FUZZY PROGRAMMING APPROACH FOR MULTI OBJECTIVE PROGRAMMING PROBLEM

Discrete Particle Swarm Optimization for TSP: Theoretical Results and Experimental Evaluations

SOLVING CAPACITATED VEHICLE ROUTING PROBLEMS WITH TIME WINDOWS BY GOAL PROGRAMMING APPROACH

For now, let us focus on a specific model of neurons. These are simplified from reality but can achieve remarkable results.

Optimum Multi DG units Placement and Sizing Based on Voltage Stability Index and PSO

Modified Seeker Optimization Algorithm for Unconstrained Optimization Problems

A PSO with Quantum Infusion Algorithm for Training Simultaneous Recurrent eural etworks

CHAPTER 7 STOCHASTIC ECONOMIC EMISSION DISPATCH-MODELED USING WEIGHTING METHOD

Optimal Solution to the Problem of Balanced Academic Curriculum Problem Using Tabu Search

A Hybrid Algorithm Based on Gravitational Search and Particle Swarm Optimization Algorithm to Solve Function Optimization Problems

Chapter Newton s Method

HYBRID FUZZY MULTI-OBJECTIVE EVOLUTIONARY ALGORITHM: A NOVEL PARETO-OPTIMIZATION TECHNIQUE

Resource Allocation with a Budget Constraint for Computing Independent Tasks in the Cloud

VARIATION OF CONSTANT SUM CONSTRAINT FOR INTEGER MODEL WITH NON UNIFORM VARIABLES

Using T.O.M to Estimate Parameter of distributions that have not Single Exponential Family

The Black Hole Clustering Algorithm Based on Membrane Computing

A destination swap scheme for multi-agent system with agent-robots in region search problems

Stochastic Weight Trade-Off Particle Swarm Optimization for Optimal Power Flow

Research Article A New Manufacturing Service Selection and Composition Method Using Improved Flower Pollination Algorithm

Economic dispatch solution using efficient heuristic search approach

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems

Heuristic Algorithm for Finding Sensitivity Analysis in Interval Solid Transportation Problems

Static security analysis of power system networks using soft computing techniques

A Multi-modulus Blind Equalization Algorithm Based on Memetic Algorithm Guo Yecai 1, 2, a, Wu Xing 1, Zhang Miaoqing 1

Thin-Walled Structures Group

Optimal Placement and Sizing of DGs in the Distribution System for Loss Minimization and Voltage Stability Improvement using CABC

Research Article A Network Traffic Prediction Model Based on Quantum-Behaved Particle Swarm Optimization Algorithm and Fuzzy Wavelet Neural Network

Conductor selection optimization in radial distribution system considering load growth using MDE algorithm

An Extended Hybrid Genetic Algorithm for Exploring a Large Search Space

Multi-Robot Formation Control Based on Leader-Follower Optimized by the IGA

Hongyi Miao, College of Science, Nanjing Forestry University, Nanjing ,China. (Received 20 June 2013, accepted 11 March 2014) I)ϕ (k)

Global Optimization Using Hybrid Approach

Dynamic Programming. Preview. Dynamic Programming. Dynamic Programming. Dynamic Programming (Example: Fibonacci Sequence)

Design of College Discrete Mathematics Based on Particle Swarm Optimization. Jiedong Chen1, a

RECENTLY, the reliable supply of electric power has been

Problem Set 9 Solutions

A new Approach for Solving Linear Ordinary Differential Equations

Extended Model of Induction Machine as Generator for Application in Optimal Induction Generator Integration in Distribution Networks

Single-Facility Scheduling over Long Time Horizons by Logic-based Benders Decomposition

THE ROBUSTNESS OF GENETIC ALGORITHMS IN SOLVING UNCONSTRAINED BUILDING OPTIMIZATION PROBLEMS

Notes on Frequency Estimation in Data Streams

Computing Correlated Equilibria in Multi-Player Games

Transcription:

1 Internatonal Congress on Informatcs, Envronment, Energy and Applcatons-IEEA 1 IPCSIT vol.38 (1) (1) IACSIT Press, Sngapore Partcle Swarm Optmzaton wth Adaptve Mutaton n Local Best of Partcles Nanda ulal Jana a*, Tapas S a and Jaya Sl b a epartment of Informaton Technology, Natonal Insttute of Technology, urgapur, Bengal, Inda b epartment of Computer Scence and Technology, BESU, West Bengal, Inda Abstract. Partcle Swarm Optmzaton (PSO) has shown ts good search ablty n many optmzaton problems. But PSO easly gets trapped nto local optma whle dealng wth complex problems due to lacks n dversty. In ths work, we proposed an mproved PSO, namely PSO-APMLB, n whch adaptve polynomal mutaton strategy s employed n local best of partcles to ntroduce dversty n the swarm space. In the frst verson of ths method (PSO-APMLB1), each local best s perturbed n the current search space nstead of entre search space. In the second verson of ths method (PSO-APMLB), each local best s perturbed n terms of the entre search space. We also proposed another local best mutaton method, namely, PSO-AMLB, n whch mutaton sze s controlled dynamcally n terms of current search space. In ths work, we carred out our experments on 8 well-known benchmark functons. Fnally the results are compared wth PSO. From the expermental results, t s found that the proposed algorthms performed better than PSO. Keywords: Partcle swarm optmzaton; adaptve polynomal mutaton;adaptve mutaton 1. Introducton Partcle swarm optmzaton [1] s a populaton based global search technque havng a stochastc nature. It has shown ts good search ablty n many optmzaton problems wth faster convergence speed. However, due to lack of dversty n populaton, PSO easly trapped nto local optma whle dealng wth complex problems. fferent mutaton strateges lke Cauchy Mutaton [3], Gaussan Mutaton [], Power Mutaton [7], Adaptve Mutaton [6] s ntroduced nto PSO for solvng local optma problem. Changhe L et al. [4] ntroduced fast partcle swarm optmzaton wth cauchy mutaton and natural selecton strategy. Andrew Stacey et al. [5] used mutaton n PSO wth probablty 1/d, where d s the dmenson of partcles. JIAOWe et al. [9] proposed elte partcle swarm optmzaton wth mutaton. Xaolng Wu et al.[7] ntroduced power mutaton nto PSO (PMPSO). Coello [13] presented a hybrd PSO algorthm that ncorporates a non-unform mutaton operator smlar to the one used n evolutonary algorthms. Pant [1] used an adaptve mutaton operator n PSO. Hgash [] proposed a PSO algorthm wth Gaussan mutaton (PSO-GM). A new adaptve mutaton by dynamcally adustng the mutaton sze n terms of current search space s proposed n [6]. A.J. Nebro et al. [1] appled a polynomal mutaton to the 15 percentage of the partcles. Tapas S et al. [17] ntroduced adaptve polynomal mutaton n global best partcles n PSO (PSO-APM). In ths work, our obectve s to use adaptve polynomal mutaton and adaptve mutaton n local best soluton n PSO to solve local optma problem and to analyss the performance and effectveness of adaptve polynomal mutaton. A comparatve study s also made wth PSO havng lnearly decreasng nerta weght.. Partcle Swarm Optmzaton(PSO) PSO s an optmzaton algorthm by smulatng the behavour of fsh schoolng and brd s flockng. PSO algorthms use a populaton of ndvdual called partcles. Each partcle has ts own poston and velocty to move around the search space. Partcles have memory and each partcle keep track of prevous best poston and correspondng ftness. The prevous best value s called as pbest. Thus pbest s related only to a

partcular partcle. It also has another value called gbest, whch s the best value of all the partcles pbest n the swarm. The basc concept of PSO technque les n acceleratng each partcle towards ts pbest and the locatons at each tme step. pbest gbest ( ) ( ) S ( t+ 1) = w v ( t) + c r x ( t) x ( t) + x ( t) x ( t) (1) 1 1 x( t+ 1) = v( t) + x( t) () c 1 and c are personal and socal cognzance of a partcles respectvely and r 1 and r are two unformly dstrbuted random numbers n the nterval (,1).The nerta weght w n (7) was ntroduced by Sh and Eberhart [14]. They proposed a w lnearly decreasng wth the teratve generaton as w= wmax ( wmax wmn ) g (3) G where g s the generaton ndex representng the current number of evolutonary generaton and G s the predefned maxmum number of generatons. 3. Proposed Algorthms Polynomal mutaton s based on polynomal probablty dstrbuton. u X = ( t+ 1) = x( t) + ( x x) δ (4) u l Where x s the upper bound and x s the lower bound of x. The parameter δϵ [-1, 1] s calculated from the polynomal probablty dstrbuton. η p( δ) =.5 ( η 1)(1 ) m m + δ (5) η m s the polynomal dstrbuton ndex. δ = { 1/( ηm+ 1) ( r) 1, r.5 1/( ηm+ 1) 1 [(1 r) ], r.5 η = 1 + t (7) m In the proposed algorthm, adaptve polynomal mutaton s done n local best of the partcles wth the mutaton probablty: 1 t 1 pm = + 1 (8) d tmax d where d s the dmenson of the problem,t s the current teraton number and s the maxmum teraton number. pbest pbest mx = x + ( b( t) a ( t)) δ (9) Where δ s calculated usng Eq.(6) and η m s calculated usng Eq.(7) a() t = mn( x()) t b() t = max( x()) t (1) In ths work, mutaton s employed n PSO wth decreasng nerta weght. The man feature of ths method s that each local best of partcles s muted wth the ncreasng probablty Pm by dynamc mutaton sze n terms of current search space and decreasng polynomal ndex wth ncreasng teraton. 3.1. PSO-APMLB1 Algorthm 1. Intalze the partcle wth unform dstrbuted random numbers.. Update velocty and poston vectors usng Eq.(1 ) and Eq.(). 3. Calculate ftness value, f ftness() 4. f ( ftness() <= Pbest()) then set Pbest() = ftness(). f (Pbest () <= gbest) then set gbest = Pbest (). 5. f (U(,1)<= Pm) then Apply mutaton to Pbest usng Eq.(8) 6. Evaluate the ftness value, Pftness() (6)

7. f ( ftness() <= Pbest()) then set Pbest() = ftness(). 8. f (Pbest () <= gbest) then set gbest = Pbest(). 9. Repeat the loop untl maxmum teraton or maxmum number of functon evaluaton s reached. In PSO-APMLB algorthm, mutaton s done by usng the followng equaton mx pbest pbest ( u = x + x x ) δ (11) In PSO-AMLB algorthm, mutaton s done by usng the followng equaton pbest pbest mx = x + ( b ( t) a ( t)) rand (1) 4. Expermental Studes There are 8 dfferent global optmzaton problems, ncludng 4 un-modal functons (f1- f4) and 4 mult-modals functons (f5- f8), are chosen n our expermental studes. These functons were used n an early study by X. Yao et al. [18]. All functons are used n ths work to be mnmzed. The descrpton of these benchmark functons and ther global optma are gven n Table 1. In ths experment, the parameters are set as followng: populaton sze s set to, =3,1 5 number of functon evaluatons are allowed for each problem. Intal dstrbuton ndex for polynomal mutaton = 1. c 1 = c = 1.49445, w max =.9 and w mn =.4 The obtaned results are presented n Tables & 3. Mean Best s the mean of best solutons and standard devaton of the best soluton of 3 runs for each test problem are reported n Table & 3.In Fg. 1, convergence graph of PSO-APLB1 for functon f5 s gven. Form the Table & 3,t can be easly sad that our proposed algorthms PSO-APMLB and PSO-AMLB performed better then Adaptve PSO.PSO-AMLB performed better for functons f 1,f 3, f 5 and f 7.PSO-APMLB performed better for functon f. PSO-APMLB1 performed better for functons f 4,f 6 and f8.for a comparson of overall performance, PSO-APMLB1 perform better for all functons except the hghly mult-modal functon f 5 comparatve to PSO-AMLB. Table 1: The 8 benchmark functons used n our experments, where s the dmenson of the functons, fmn s the mnmum values of the functons, and S R n the search space Test Functon S f mn f 1( x) = x = 1 f ( x ) = x = 1 = 1 1 6 n 1 3 ( ) = ( 1 ) = 1 f x x 5. Concluson 1 4( ) = [1( + 1 ) + ( 1) ] = 1 5( ) = * sn( ) = 1 f x x x x f x x x x x f 6( x) = cos( ) + 1 = 1 4 f 1 1 7 ( x ) = * exp(. * x ) exp( cos( π x )) + + e f 8 x x π x = 1 = 1 = 1 ( ) = [ 1 cos( )) + 1] [-1,1] [-1,1] [-1,1] [-1,1] [-5,5] -1569.5 [-6,6] [-3,3] [-5.1,5.1] In ths work, we proposed adaptve polynomal mutaton and adaptve mutaton n local best of partcles n partcle swarm op-best of partcles to ntroduce dversty n the populaton and to solve local optma problem. PSO wth adaptve polynomal mutaton and adaptve mutaton n local best of partcles produces the better results than PSO. But PSO-APMLB produces poor performance for mult-modal functon f 5 and f 8

wth respect to the global optmums of the aforementoned functons. Our future works wll be drected towards solvng local mnma problem n complex mult-modal functon optmzaton. Fg. 1: Convergence graph of PSO for f5 Table : Functon values acheved by PSO and PSO-APMLB1 PSO wth decreasng ω PSO-APMLB1 Problem Mean Std. ev. Mean Std. ev. f 1.58e-11 1.37e-1 1.5e-13 5.34e-13 f 8.54 9.5 3.84 5.1 f 3 1.31e-1.45-1 5.41e-1.16e-9 f 4 51.7 3.3 1.9 4.31 f 5-91.46 53.14-149.3 759.474 f 6 1.81e-.41e- 1.54e- 1.5e- f 7.566.7.464.753 f 8 51. 15.4 49.35 14 Table 3: Functon values acheved by PSO-APMLB and PSO-APML 6. References Problem PSO-APMLB PSO-AMLB Mean Std. ev. Mean Std. ev. f 1.64E-13 1.31E-1 3.7e-15 1.68e-14 f.169.154 6.68 1. f 3 3.36e-9 1.56e-8 1.86e-14 5.93e-14 f 4 6.5 1.6 51. 39. f 5-98.57 6.96-131.68 5.9641 f 6.54e-.5e- 1.76e- 1.77e- f 7.746.835.61.36 f 8 51.4 14.5 91. 4.1 [1] Kennedy J and Eberhart R. Partcle Swarm Optmzaton, In Proc. IEEE Internatonal Conference on Neural Networks, Perth, Australa, 1995. [] Hgash N, lba H, Partcle Swarm Optmzaton wth Gaussan Mutaton, In Proc. IEEE Swarm Intellgence Symposum, Indanapols, 3, pp. 7-79 [3] Wang H, Lu Y, L C. H., Zeng S. Y. A hybrd partcle swarm algorthm wth Cauchy mutaton, In Proc. of IEEE Swarm Intellgence Symposum, 7, pp. 356-36.

[4] L C, Lu Y., Zhao A, Kang L, Wang H., A Fast Partcle Swarm Algorthm wth Cauchy Mutaton and Natural Selecton Strategy,,LNCS 4683, 7, pp. 334-343, Sprnger-Verlag [5] Stacey A, Jancc M, Grundy I,Partcle swarm optmzaton wth mutaton, n Proc. Congr. Evol. Comput, 3, pp. 145 143 [6] Tang J, Zhao X, Partcle Swarm Optmzaton wth Adaptve Mutaton, WASE Internatonal Conference on Informaton Engneerng,9 [7] Wu X, Zhong M, Partcle Swarm Optmzaton Based on Power Mutaton, ISECS Internatonal Colloquum on Computng, Communcaton, Control, and Management,9 [8] Gao Y, Ren Z, Adaptve Partcle Swarm Optmzaton Algorthm Wth Genetc Mutaton Operaton,Thrd Internatonal Conference on Natural Computaton(ICNC 7). [9] We J, Guangbn L,ong L, Elte Partcle Swarm Optmzaton wth Mutaton, In Proc. Asa Smulaton Conference - 7th Internatonal Conference on System Smulaton and Scentfc Computng, 8. ICSC 8. [1] Nebro A.J, SMPSO: A New PSO-based Metaheurstc for Mult-obectve Optmzaton, eee symposum on Computatonal ntellgence n mult-crtera decson-makng,9 [11] eb K, Mult-Obectve Optmzaton Usng Evolutonary Algorthms,John Wley and Sons,1. [1] Pant M, Thangara R, Abraham A., Partcle Swarm Optmzaton usng Adaptve Mutaton, n Proc. 19th Internatonal Conference on atabase and Expert Systems Applcaton, 8, pp. 519-53 [13] Esquvel S. C, Coello Coello C. A., On the use of partcle swarm optmzaton wth multmodal functons,n Proc. Congr. Evol.Comput., 3, pp. 113-1136, [14] Chen J, Ren Z, Fan X, Partcle swarm optmzaton wth adaptve mutaton and ts applcaton research n tunng of PI parameters, n Proc. 1st Internatonal Symposum on Systems and Control n Aerospace and Astronautcs, 6, pp. 99-994 [15] Sh Y, Eberhart R.C. A modfed partcle swarm optmzer, n proc. IEEE World Congr. Comput.Intell.,1998,pp.69-73 [16] X. Yao, Y. Lu and G. Ln, Evolutonary programmng made faster,ieee Transactons on Evolutonary Computaton.,vol. 3,1999,pp. 8-1 [17] S T, Jana N., Sl J. Partcle Swarm Optmzaton wth Adaptve Polynomal Mutaton, n proc. World Congress on Informaton and Communcaton Technologes(WICT 11),Mumba,Inda,11,pp. 143 147