Is there a way to exploit symmetry in black box optimization? Specifically, I want to find a local minimum of a function $f: \mathbb{R}^{300} \rightarrow \mathbb{R}$ which has the property that any permutation of the inputs gives the same output. Apart from that, $f$ doesn't have any "nice" property that I know of (it's certainly not smooth).
I have tried a number of black-box optimization techniques (Nelder-Mead, COBYLA, Bayesian optimization...) but they take a prohibitively long time and haven't produced satisfying results. The symmetry in the function's inputs makes me think there should be a way to reduce the impact of the "curse of dimensionality".
Any idea is welcome!