Currently, agent novelty is calculated with respect to the parent population or the child population, but it should actually be computed with respect to the combined parent/child population, at least right before selection is performed on the combined population. I suppose that we still need parent population novelty before children are generated.
Currently, agent novelty is calculated with respect to the parent population or the child population, but it should actually be computed with respect to the combined parent/child population, at least right before selection is performed on the combined population. I suppose that we still need parent population novelty before children are generated.