Multistart lsqcurvefit with parallel processing on produces an error

2 次查看(过去 30 天)
so I have a lsqcurvefit problem defined with a function and I run it with MultiStart to get a global minimum. Everything works perfectly fine if I have "useparallel" disabled. problem=createOptimProblem('lsqcurvefit','x0',x0,'objective',@myfunction,... 'lb',lb,'ub',ub,'xdata',xdata,'ydata',ydata,'options',options);
However, when I try parallel processing it gives me the error: "Attempted to access x0(1); index out of bounds because numel(x0)=0." x0 contains the parameters of my function to fit (which is a separate .m file). Apparently the input variables of my function are not available in the matlab workers. I tried to make x0 a global variable, but then I have the problem that the local values during fitting keep getting changed back to the global start value...
  3 个评论
Felix
Felix 2013-3-31
I'm sure its not an empty vector (x0 = [0.1 2 3 50];), and its passed into the function to fit as input argument. Since I get nice results with Multistart when I have "useparallel" disabled, I don't think there is any programming error. It's just another curiosity about the parallel processing toolbox, maybe it doesn't accept that I have my fit-function in a separate m-file instead of having it as an anonymous inline function. I could imagine that it gets confused to run the function file multiple times with multiple values for the same variable name x0....
Matt J
Matt J 2013-3-31
I could imagine that it gets confused to run the function file multiple times with multiple values for the same variable name x0
I can't imagine that. It's the whole point of MultiStart. You'll have to show more code or we'll just be guessing.

请先登录,再进行评论。

回答(0 个)

类别

Help CenterFile Exchange 中查找有关 Global or Multiple Starting Point Search 的更多信息

产品

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by