Is NGSA-II deterministic?

[dinohsu1019]

I always use the same seed and expect the result to be deterministic.

algorithm.setup(
    problem,
    termination = termination,
    seed = 1,
    verbose = True,
    save_history = True
)

And I fix one of the solution variables, and performance is significantly less as good, although I assume that solution variable to of minor impact to the result. Another factor is the underlined data can change over time, so the different runs get different results on the same computer, and I assume that should be minor too.

Anyway, back to the fundamental question: is this algorithm deterministic among two runs on the same computer with the same seed?

A peer quant trader says the "better result" should be always the same, as long as the algorithm goes long enough. I have a custom termination criteria in addition to the max generations, say, 400, I have it terminate if the objectives have no improvement for, say, 60, contiguous generations, and the 60 is discretionary, is it too short? Another thing to consider is the pop size now is 80 and replacement across generations is only 40, do I have to increase pop size or increase, decrease replacement solutions to allow more exploration?

Thanks in advance.