ES fails in the _infill() method, when duplicates are removed from the population

[ernstklrb]

pymoo/pymoo/algorithms/soo/nonconvex/es.py

Line 55 in 70b8ef0

n_offsprings=n_offsprings,

at the above line, the argument
eliminate_duplicates=False,

should be added.

Because, when duplicates are removed in the course of the algorithm running, the actual population size will become less than the value in self.pop_size. This, in turn, will lead to the index I in _infill() going beyond the size of the array X, which results in an error.

(This happened to me when I initialized the population with all zero values to see how well the ES would recover from all solutions being infeasible in my problem - that went well (with the above fix)!)

[ernstklrb]

additional remark: I think that - independent of that concrete problem in _infill() - removing duplicates does not make sense in general for ES: if a solution is / stays in the population in some multiplicity, this is an indicator of the solution being of some value, and making them more "present" in the offspring