Global optimization hybrid gamultiobj

Hello

I am using hybrid gamultiobj() to solve a optimization problem. As of now i am trying to reduce the execution time as much as possible without using parallelcomputing toolbox.

Hence i tried the below steps:

• Call gamulitobj()
• Call OutputFcns() - In this function, i am monitoring the intermediate states and calling a custom made decision making algorithm, which selects one optimal parameter from the set of optimal solutions with rank 1 at each iteration
• Based on a custom made condition after analyzing past history, i terminate the global optimization using 'state.StopFlag'
• Just before terminating the global optimization, I overwrite the state.Population, state.Score,state.Rank with only one optimal parameter.

   state.Population = Top_Population; %decided by a utility function
   state.Score  = Top_Score; %decided by a utility function
   state.Rank = 1; % Only one optimal parameter of rank 1.

• Later algorithm switches over to a multiobjective local optimizer (fgoalattain). However when i monitor the iterations of fgoalattain using OutputFcns_hyb() i always get 'Hessian not updated'

I am sure that something is going wrong at this stage.

Bascially, I understand that 'fgoalattain' runs for all pareto optimal solutions. In my case i wish to run it for only one optimal solution that is with the return value (optimal parameter) of my custom made decision making algorithm.

In summary,

• Can we overwrite the state structure as described above?
• If i cannot overwrite, can i use gamultiobj() without hybrid function and later call fgoalattain with the single optimal parameter as initial condition?
• Any simple example to call fgoalattain() as a stand alone function will be really helpful as i could not clearly understand the 'GOAL' and 'WEIGHT' input parameters. Any default setting can be used for GOAL and WEIGHT?

Thank you

Chris

[christober]

Instead of overwriting the state structure, i called the fgoal attain with single optimal parameter as initial condition.

        .....
        Scaled_Init =[4.2743,0.8293,3.4810,4.7922,137.2542,0.6860,0.3834]; %Single optimal parameters [output of terminated global optimizer] 
        goal = [0.0661,0.1176]; %fval of single optimal parameters[Two objectives]
        weight = abs(goal);
        f=@(x)multiobjopt(x,Exp_Data,model,numel(d));
        options = optimoptions(@fgoalattain,'Display','iter');
        [x,fval,attainfactor,exitflag,output]= fgoalattain(f,Scaled_Init,goal,weight,[],[],[],[],LB,UB,[],options);

By doing it this way i avoided the execution of local optimizer for all the set of optimal solutions.