FIBO directly samples the optimum from a pretrained in-context generative model, bypassing surrogate fitting and acquisition function maximization, and matches GP-based Bayesian optimization in quality while being far faster.
Using confidence bounds for exploitation-exploration trade-offs.JMLR, 3(null), 2003
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Simplifying Bayesian Optimization Via In-Context Direct Optimum Sampling
FIBO directly samples the optimum from a pretrained in-context generative model, bypassing surrogate fitting and acquisition function maximization, and matches GP-based Bayesian optimization in quality while being far faster.