Prev: Color Order Issue
Next: checkboxes in word documents
From: Ritesh on 14 Jul 2010 14:14
Hi,

I am trying to implement nelder mead algorithm to an optimization problem usng derivative free approach. But the problem is I have a large number of parameters (around ten thousand). So to use this I have to to create a simplex of "number of parameters + 1" sides and evaluate function value at each point. This is very costly as function evaluation take around 10 sec. I have used projected gradient algorithm for the same problem and is trying to develop some quicker algorithm for this specific problem. Is it possible to use Nelder Mead with some variation? If not, is there any other derivative free method that can fasten up the process in this particular case?

Thanks
From: Ritesh on 15 Jul 2010 08:19
Hi,

I am trying to implement nelder mead algorithm to an optimization problem usng derivative free approach. But the problem is I have a large number of parameters (around ten thousand). So to use this I have to to create a simplex of "number of parameters + 1" sides and evaluate function value at each point. This is very costly as function evaluation take around 10 sec. I have used projected gradient algorithm for the same problem and is trying to develop some quicker algorithm for this specific problem. Is it possible to use Nelder Mead with some variation? If not, is there any other derivative free method that can fasten up the process in this particular case?

Thanks
 | 
Pages: 1
Prev: Color Order Issue
Next: checkboxes in word documents