How can I come to know that the Constraints are satisfied in Pattern Search algorithm?

Hi I am using Patternsearch algorithm to find minimum value of objective function. Pattern Search algorithm terminates when number of generations exceeds. How can I come to know that the Constraints are satisfied?

Answer 1

[Walter Roberson]

You can set the options to display information for each iteration. Or you can use the four-output variation and look at the output structure for maxconstraint

https://www.mathworks.com/help/gads/patternsearch.html#outputarg_output

[Dhanu Naik]

I am getting maxconstraints=1.2034e-07 What is the meaning of this?

[Walter Roberson]

You have a violated constraint but probably just slightly violated. For example a value might be at a value where you programmed that it needed to be inside the boundary. Or you might be having floating point roundoff errors.

But sometimes a small value for violated constraints can still point to an unacceptable configuration. For example I once programmed a ga search that required that the values were in a particular order, and the best the search was able to do after quite a while was to get everything else right except that two values were exchanged. From the perspective of the goal of that particular search, it was a hard failure to have the values out of order. But the best it came up with might help giude human analysis.

[Dhanu Naik]

Do I need to vary other parameters of Patternsearch algorithm untill maxconstraints approaches Zero?

[Walter Roberson]

I would be more likely to loop doing patternsearch with a limited number of iterations but different x0 vectors, hoping to find locations where the constraints are satisfied. This is basically the same technique used by multiStart() but unfortunately multiStart does not appear to support patternsearch.