PSO源程序.docx

1、PSO源程序/PSO头文件PSO.cpp如下:/Header _PSO_H/n/PlatfoPurpose Provide a class for particle swarm optimizatiorm Visual C+ 7.0/群微粒算法：本算法用群微粒算法求目标函数的最大值/本算法使用步骤/ （1）派生自己的群微粒类，类中必须定义double GetFit(PARTICLE&)方法，用来计算每个微粒的适合度/ （2）生成派生类实例，并在构造函数中指明微粒坐标维数和群体个数/ （2）设置微粒坐标上界数组和下界数组，并用SetXup与SetXdown设置微粒坐标上下界/ （3）用SetVm

2、ax方法设置微粒最大速度/ （4）设置可选参数：C1，C2，W和通讯函数/ （5）采用Run方法进行优化运算，优化后用GetBest方法获得最优个体适合度和坐标#ifndef _PSO_H#define _PSO_H/微粒类class PARTICLEpublic: double *X; /微粒的坐标数组 double *V; /微粒的速度数组 double *XBest; /微粒的最好位置数组 int Dim; /微粒的维数 double Fit; /微粒适合度 double FitBest; /微粒最好位置适合度 /构造函数 PARTICLE(); /空构造函数 PARTICLE(int

3、n); /维数为参数的构造函数 /析构函数 PARTICLE(); void SetDim(int d); /设置微粒的维数;/定义群粒子类class PSOprotected: PARTICLE *Particle; /微粒群数组 int PNum; /微粒个数 int GBestIndex; /最好微粒索引 double W; /惯性权重 double C1; /加速度系数1 double C2; /加速度系数2 double *Xup; /微粒坐标上界数组 double *Xdown; /微粒坐标下界数组 double *Vmax; /微粒最大速度数组 void Initialize()

4、; /初始化群体 void CalFit(); /计算全体适合度 virtual void ParticleFly(); /微粒飞翔，产生新一代微粒 /通讯函数，返回值为false时，系统停止优化 bool (*Com)(double, /最优微粒适合度 double*, /最优微粒坐标数组 double*, /所有微粒坐标指针数组 int); /当前最优微粒索引public: /构造函数 PSO(); /空构造函数 PSO(int dim, /微粒维数 int num); /微粒个数 /析构函数 PSO(); void SetXup(double*); /设置微粒坐标上界 void SetX

5、down(double*); /设置微粒坐标下界 void SetVmax(double*); /设置微粒最大速度，以数组为参数 void SetVmax(double); /设置微粒最大速度，以上下界百分比为参数 void SetW(double w)W=w; /设置权重 void SetC1(double c)C1=c;/设置C1 void SetC2(double c)C2=c;/设置C2 void SetCom(void *p)Com=(bool(*)(double,double*,double*,int)p;/设置通讯函数 /计算特定微粒坐标所对应适合度，必须由派生的实际PSO类定义

6、，以便计算适合度 virtual double GetFit(PARTICLE&)=0; /运行类进行优化 PARTICLE& Run(int max); /按最多次数限制运行PSO PARTICLE& Run(double fit); /按最佳适合度目标运行PSO double GetBest(double*); /获得最优微粒适合度和坐标;#endif/PSO源文件PSO.CPP如下:/Header _PSO_H/Purpose Provide a class for particle swarm optimization/Platform Visual C+ 7.0/Date 2002.

7、4.29/Author Liu Kang/群微粒算法：本算法涌群微粒算法求目标函数的最大值/本算法使用步骤/ （1）派生自己的群微粒类，类中必须定义double GetFit(PARTICLE&)方法，用来计算每个微粒的适合度/ （2）生成派生类实例，并在构造函数中指明微粒坐标维数和群体个数/ （2）设置微粒坐标上界数组和下界数组，并用SetXup与SetXdown设置微粒坐标上下界/ （3）用SetVmax方法设置微粒最大速度/ （4）设置可选参数：C1，C2，W和通讯函数/ （5）采用Run方法进行优化运算，优化后用GetBest方法获得最优个体适合度和坐标#include #includ

8、e #include /#include PSO.H/微粒构造函数PARTICLE:PARTICLE() /空构造函数 X = 0; V = 0; XBest = 0; Dim = 0;PARTICLE:PARTICLE(int n) /维数为参数的构造函数 Dim = n; X = new doubleDim; V = new doubleDim; XBest = new doubleDim;/微粒析构函数PARTICLE:PARTICLE() if(Dim) delete X; delete V; delete XBest; /设置微粒的维数void PARTICLE:SetDim(int

9、 d) if(X) delete X; if(V) delete V; if(XBest) delete XBest; Dim = d; X = new doubleDim; V = new doubleDim; XBest = new doubleDim;/PSO构造函数PSO:PSO() Particle = 0; PNum = 0; GBestIndex = 0; Xup = 0; Xdown = 0; W = 1; C1 = 2; C2 = 2; Com = 0;PSO:PSO(int dim, int num) Particle = new PARTICLEnum; for(int

10、i=0; inum; i+) Particlei.SetDim(dim); PNum = num; GBestIndex = 0; Xup = new doubledim; Xdown = new doubledim; Vmax = new doubledim; W = 1; C1 = 2; C2 = 2; Com = 0;/析构函数PSO:PSO() if(Particle) delete Particle; if(Xup) delete Xup; if(Xdown) delete Xdown; if(Vmax) delete Vmax;/设置坐标上界void PSO:SetXup(doub

11、le *up) if(!Particle) return; for(int i=0; iParticle0.Dim; i+) Xupi = upi;/设置坐标下界void PSO:SetXdown(double *d) if(!Particle) return; for(int i=0; iParticle0.Dim;i+) Xdowni = di;/设置最大速度void PSO:SetVmax(double *max) if(!Particle) return; for(int i=0; iParticle0.Dim;i+) Vmaxi = maxi;void PSO:SetVmax(dou

12、ble p) if(!Particle) return; for(int i=0; iParticle0.Dim; i+) Vmaxi = (Xupi-Xdowni)*p;/初始化群体void PSO:Initialize() if(!Particle) return; static int kk=(unsigned)time(NULL); srand(unsigned)time(NULL)+kk+); GBestIndex = 0; for(int i=0; iPNum; i+)/初始化所有的粒子个体 for(int j=0; jParticleGBestIndex.Fit) GBestIn

13、dex = i;/如果这个鸟的适合度大于群体的最大适合度的话，记录下来查找群体最优微粒 /计算群体各个微粒适合度void PSO:CalFit() if(!Particle) return; for(int i=0; iPNum; i+) Particlei.Fit = GetFit(Particlei);/微粒飞翔,产生新一代微粒void PSO:ParticleFly() static double FitBak100;/用来存放备份的合适度值 if(!Particle) return; static int tt=(unsigned)time(NULL);/wo:以时间为产生随机数的种子

14、 srand(unsigned)time(NULL)*tt+);/wo:srand()的作用是为rand()设置种子,两者配合使用的 /整个群体飞向新的位置 for(int i=0; iPNum; i+) for(int j=0; jParticlei.Dim; j+) Particlei.Vj = W*Particlei.Vj+/ (FitBaki-Particlei.Fit)+/修改速度 rand()/(double)RAND_MAX*C1*(Particlei.XBestj-Particlei.Xj)+ /wo:此处的产生的两个rand()是一样的，有待改进? /wo:RAND_MAX为

15、最大随机数 rand()/(double)RAND_MAX*C2*(ParticleGBestIndex.XBestj-Particlei.Xj); for(j=0; jVmaxj) Particlei.Vj = Vmaxj; if(Particlei.Vj-Vmaxj) Particlei.Vj = -Vmaxj; for(j=0; jXupj) Particlei.Xj=Xupj;/保护 if(Particlei.XjXdownj) Particlei.Xj=Xdownj; /计算各微粒适合度 CalFit(); for(i=0; iPNum; i+) FitBaki = Particle

16、i.Fit; /设置新的个体最好位置 for(i=0; i=Particlei.FitBest) Particlei.FitBest = Particlei.Fit; for(int j=0; jParticlei.Dim; j+) Particlei.XBestj = Particlei.Xj; /设置群体中新的最优个体 GBestIndex = 0; for(i=0; i=ParticleGBestIndex.FitBest & i!=GBestIndex) GBestIndex = i;/运行群粒算法PARTICLE& PSO:Run(int n) Initialize(); doubl

17、e *opt_p = new doubleParticle0.Dim; /通讯用数组，最优点坐标 double *opt_a = new double*PNum; /通讯用数组，所有点坐标 for(int i=0; in; i+) ParticleFly(); if(Com) /通讯函数存在，完成通讯 for(int k=0; kParticle0.Dim; k+) opt_pk=ParticleGBestIndex.XBestk;/拷贝最优点坐标 for(k=0; kPNum; k+) opt_ak = Particlek.X; /指向所有点坐标 if(!Com(ParticleGBestI

18、ndex.FitBest,opt_p,opt_a,GBestIndex) break; delete opt_p; delete opt_a; return ParticleGBestIndex;PARTICLE& PSO:Run(double fit) double *opt_p = new doubleParticle0.Dim; /通讯用数组，最优点坐标 double *opt_a = new double*PNum; /通讯用数组，所有点坐标 Initialize(); do ParticleFly(); if(Com) /通讯函数存在，完成通讯 for(int k=0; kParti

19、cle0.Dim; k+) opt_pk=ParticleGBestIndex.XBestk;/拷贝最优点坐标 for(k=0; kPNum; k+) opt_ak = Particlek.X; /指向所有点坐标 if(!Com(ParticleGBestIndex.FitBest,opt_p,opt_a,GBestIndex) break; while(ParticleGBestIndex.FitBestfit); delete opt_p; delete opt_a; return ParticleGBestIndex;/返回最佳个体double PSO:GetBest(double *r

20、) for(int i=0; i 1000#pragma once#endif / _MSC_VER 1000#define WIN32_LEAN_AND_MEAN / Exclude rarely-used stuff from Windows headers#include / TODO: reference additional headers your program requires here/AFX_INSERT_LOCATION/ Microsoft Visual C+ will insert additional declarations immediately before

21、the previous line.#endif / !defined(AFX_STDAFX_H_CA78CC4A_4FB0_49B9_B823_A4DBAEFB44D2_INCLUDED_)*#ifndef _LIBSVM_H#define _LIBSVM_H#ifdef _cplusplusextern C #endifstruct svm_node int index; double value;struct svm_problem int l; double *y; struct svm_node *x;enum C_SVC, NU_SVC, ONE_CLASS, EPSILON_SV

22、R, NU_SVR ; /* svm_type */enum LINEAR, POLY, RBF, SIGMOID, PRECOMPUTED, MIX,COMBINATION ; /* kernel_type */struct svm_parameter int svm_type; int kernel_type; int degree; /* for poly */ double gamma; /* for poly/rbf/sigmoid */ double coef0; /* for poly/sigmoid */ double t; /*for mix/combination*/ /*

23、 these are for training only */ double cache_size; /* in MB */ double eps; /* stopping criteria */ double C; /* for C_SVC, EPSILON_SVR and NU_SVR */ int nr_weight; /* for C_SVC */ int *weight_label; /* for C_SVC */ double* weight; /* for C_SVC */ double nu; /* for NU_SVC, ONE_CLASS, and NU_SVR */ do

24、uble p; /* for EPSILON_SVR */ int shrinking; /* use the shrinking heuristics */ int probability; /* do probability estimates */;struct svm_model *svm_train(const struct svm_problem *prob, const struct svm_parameter *param);void svm_cross_validation(const struct svm_problem *prob, const struct svm_parameter *param, int nr_fold, double *target);int svm_save_model(const char *model_file_name, const struct svm_model *model);struct svm_model *svm_load_model(const char *model_file_name);int svm_get_svm_type(const struct svm_model *model);int svm_get_nr_class(co