1,000 nds all in the same location

[aaronlrees]

Hello,

I'm trying to minimise the output of a Gaussian process using NSGA2. This has two outputs: prediction and uncertainty. Here I'm negating uncertainty to minimise prediction whilst maximising uncertainty. My question is regarding the results that I'm getting: what are some possible reasons for having 1,000 nds all in the same location? - is it due to an error in the code?

See the code below and output for when verbose=True, along with the plot of the nds.

Thank you

Code:

  class MyProblem(Problem):
      def __init__(self, dim, ub, lb, model):
          super().__init__(
              n_var=dim, n_obj=2, n_constr=0, xl=lb, xu=ub
          )
          self.model = model
  
      def _evaluate(self, x, out, *args, **kwargs):
          mu, std = self.model._predict(x)
          f1 = mu
          f2 = -std

          out["F"] = np.column_stack([f1, f2])
  
  
  problem = MyProblem(len(ub), ub, lb, model)
  
  mutation = PolynomialMutation(prob = 1/len(ub), eta=20)
  crossover = SBX(prob_var=0.8, eta=20)
  
  algorithm = NSGA2(pop_size=100 * len(ub),
                    crossover=crossover,
                    mutation=mutation,
                    sampling=LHS(),
                  eliminate_duplicates=True)
  termination = get_termination("n_gen", n_max_gen=50) 
  
  res = minimize(problem,
         algorithm,
         termination,
         seed=1,
         verbose=True)

Outputs:

Thank you

[aaronlrees]

Solved it. It is not a pymoo issue but a GP kernel issue. My bad! Just wondering though... does everything look as it should above in terms of the code? Thank you