优化方法上机作业.docVIP

第一题（牛顿法和不精确一维搜索） 牛顿法： syms x1 x2; f=(x2-(x1)*(x1))^2+(1-x1)^2; v=[x1,x2]; df=jacobian(f,v); df=df.; G=jacobian(df,v); epson=1e-12;x0=[0,1];g1=subs(df,{x1,x2},{x0(1,1),x0(2,1)});G1=subs(G,{x1,x2},{x0(1,1),x0(2,1)});k=0;mul_count=0;sum_count=0; mul_count=mul_count+12;sum_count=sum_count+6; while(norm(g1)epson) p=-G1\g1; x0=x0+p; g1=subs(df,{x1,x2},{x0(1,1),x0(2,1)}); G1=subs(G,{x1,x2},{x0(1,1),x0(2,1)}); k=k+1; mul_count=mul_count+16;sum_count=sum_count+11; end; k x0 mul_count sum_count 运行结果： k = 9 x0 = 1 1 mul_count = 156 sum_count = 105 不精确一维搜索： fun1.m文件 function f=fun1(x) f=(x(1)-1)^2+(x(2)-x(1)^2)^2; gfun1.m文件 function gf = gfun1(x) gf=[2*x(1)*(x(1)^2-x(2))+2*(x(1)-1),2*(x(2)-x(1)^2)]; wolfepowell.m文件 function [k,m,opt,x]=wolfepowell(xk,sk) max=1000; c1=0.1; c2=0.6;a=0;b=inf;dk=0.2;m=0; while(m=max) if (fun1(xk)-fun1(xk+sk*dk)-c1*dk*gfun1(xk)*sk) b=dk;dk=(dk+a)/2; elseif (fun1(xk)-fun1(xk+sk*dk)=-c1*dk*gfun1(xk)*sk)(gfun1(xk+sk*dk)*skc2*gfun1(xk)*sk); a=dk; if 2*dk(dk+b)/2 dk=2*dk; else dk=(dk+b)/2; end else mk=m; break; end m=m+1; end m=mk;k=dk;x=xk+sk*dk;opt=fun1(x); 运行结果： xk=[0,1];sk=[1,-1]; [k,m,opt,x]= wolfepowell(xk,sk) k = 0.8000 m = 2 opt = 1.0080 x = 0.8000 0.2000 第二题（共轭梯度法） frcg.m文件 function [x,val,k]=frcg(fun,gfun,x0) maxk=5000; rho=0.6;sigma=0.4; k=0; epsilon=1e-4; n=length(x0); while(kmaxk) g=feval(gfun,x0); itern=k-(n+1)*floor(k/(n+1)); itern=itern+1; if(itern==1) d=-g; else beta=(g*g)/(g0*g0); d=-g+beta*d0; gd=g*d; if(gd=0.0) d=-g; end end if(norm(g)epsilon), break; end m=0; mk=0; while(m20) if(feval(fun,x0+rho^m*d)feval(fun,x0)+sigma*rho^m*g*d) mk=m; break; end m=m+1; end x0=x0+rho^mk*d; val=feval(fun,x0); g0=g; d0=d; k=k+1; end x=x0