Can I avoid this slow for loop?

Question

Filippo 2015-8-21

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/235118-can-i-avoid-this-slow-for-loop

关闭： MATLAB Answer Bot 2021-8-20

I am writing a program to compute a value function which is defined over a 3 dimensional state space. It's a dynamic programming problem which I have managed to reduce to 2 control variables. The code works, but every iteration takes about 1 minute and it takes about 100 iterations to reach convergence. Is there a simple way to get rid of the "for" loop and vectorize also that part of the code? Thanks

%STATE SPACE;
mu=(0.01:0.01:.9);
b=-3:.05:0;
delta=0.01:0.05:1.2;
[DELTA,B,MU]=meshgrid(delta,b,mu); 
iter=1;
norm=1001;
tol=.001;
% Initial Guess for Value Function
V0=0.*DELTA;
%Control variables grid
y=[ya:0.05:yb];
[CA,CB]=meshgrid(y,y);
iter=1;
norm=1001;
tol=.01;
%Main loop
while norm>tol
  for i=1:length(mu)
    for j=1:length(b)
      for z=1:length(delta)
        dp=f(mu(i),b(j),delta(z),CB,CA);
        bp=g(mu(i),b(j),delta(z),CB,CA);
        mup=h(mu(i),b(j),delta(z),CB,CA);
        fval= l(mu(i),CA,CB)+beta*interp3(DELTA,B,MU,V0,dp,bp,mup,'linear');
       [mr mc]=find(fval==max(max(fval)));
        V1(j,z,i)=fval(mr(1),mc(1));
      end
    end
  end
  toc
  norm(iter)=max(max(max(abs(V0-V1))))
  V0=V1;
  iter=iter+1
end

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再回答此问题。

Answer 1

Varun Bhaskar 2015-8-25

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/235118-can-i-avoid-this-slow-for-loop#answer_190318

Hi Filippo,

You can parallelize your code to be executed on different cores on your machine. You can find more information about the 'parfor' loop construct in the following link :

Parfor Loops

Parfor Limitations