Across 17 benchmark and real-world tasks, non-GP surrogates (RF, NGBoost, BASS) match or beat Gaussian-process BO while using a fraction of the compute and memory, and a cheap-feature classifier can predict the best surrogate.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Frugal Bayesian Optimization: Scalable Surrogates for Data- and Resource-Limited Discovery
Across 17 benchmark and real-world tasks, non-GP surrogates (RF, NGBoost, BASS) match or beat Gaussian-process BO while using a fraction of the compute and memory, and a cheap-feature classifier can predict the best surrogate.