Setting an initial population with a known objective function value

[isong29]

Hi! I'm having some trouble setting up a GA optimization when I want to give it an initial population. I read from https://pymoo.org/customization/initialization.html that I can have a pre-evaluated population. However, the sample code calls Evaluator().eval(problem, pop). What if I already have an excel file, say, with all of the X values with the corresponding F values?

I tried to create a population with known F values, something like:
pop = Population.new("X", X, "F", F)
where X is a 100 sample parameter sets, and F is a 100 sample objective function values.

However, if I insert this pop directly into the algorithm, it evaluates each parameter set, despite already having values of "F". Is there a way to bypass this? In other words; following the example on the website, I would like to do something like this:

X = np.random.random((300, problem.n_var))
F = np.random.random((300, 1))
pop = Population.new("X", X, "F", F)

algorithm = NSGA2(pop_size=100, sampling=pop)

minimize(problem,
algorithm,
('n_gen', 10),
seed=1,
verbose=True)

And I would like to have the algorithm skip evaluating the initial 300 samples, and take the "F" values that already exist in the population object. Is that possible?

Thank you!

[isong29]

Actually I may have just found the answer; do I just have to evaluate using a StaticProblem?

[blankjul]

yes StaticProblem is one solution.
The main reason why it is evaluated again is that more attributes need to be set, e.g. constraint violation, ....