Hi Saurabh,
As per my understanding you want to use four trained Gaussian process regression models with six input variables as objective functions in a multiobjective optimization problem in MATLAB. For this purpose, you need to create a function that takes the decision variables as input and returns the predictions of the GPR models as the objectives as shown below:
function objectives = myObjectiveFunction(X, modell1, modell2, modell3, modell4)
% Predictions from the trained GPR models passed as parameters
obj1 = predict(modell1, X);
obj2 = -predict(modell2, X); % Negate to maximize
obj3 = predict(modell3, X);
obj4 = predict(modell4, X);
objectives = [obj1, obj2, obj3, obj4];
end
Since you want to minimize three of the outputs and maximize one, you should negate the output of the model you wish to maximize. This function can then be passed to an optimizer that supports multiobjective optimization such as 'gamultiobj' in MATLAB.
You can modify your code after defining the objective function as shown below:
% Create an anonymous function that captures the trained models and passes them to the objective function
objectiveWithModels = @(X) myObjectiveFunction(X, modell1, modell2, modell3, modell4);
% Run the multiobjective optimization using the anonymous function
options = optimoptions('gamultiobj', 'PlotFcn', @gaplotpareto, 'PopulationSize', 50, 'MaxGenerations', 30);
[x, fval] = gamultiobj(objectiveWithModels, 6, A, b, Aeq, beq, lb, ub, options);
You can refer to the following documentation link to read more about 'gamultiobj' function:
Hope it helps.
