A generalization of probabilistic reparameterization allows gradient-based acquisition optimization in fully mixed-variable Bayesian optimization with Gaussian process surrogates for non-equidistant discrete spaces.
URL https://link.aps.org/doi/10.1103/ PhysRevLett.124.086102
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Bayesian Optimization for Mixed-Variable Problems in the Natural Sciences
A generalization of probabilistic reparameterization allows gradient-based acquisition optimization in fully mixed-variable Bayesian optimization with Gaussian process surrogates for non-equidistant discrete spaces.