BO-pro-c-tr
An algorithm for large-scale Bayesian optimization that combines a product-of-experts Gaussian process with uncertainty calibration (GP-pro-c) and a one-to-many trust-region association strategy. The algorithm is designed to address the boundary discontinuity and low computational efficiency of conventional independent local Gaussian process models.
Timeline (last 90 days)
The paper proposes an algorithm named BO-pro-c-tr that uses a product-of-experts Gaussian process with uncertainty calibration (GP-pro-c) to重构 the paradigm of large-scale Bayesian optimization.
When combined with the GP-LCB acquisition function, BO-pro-c-tr reduces average simple regret by 2.6% and computational overhead by 34.9% relative to methods based on independent local GPs and trust regions.
The paper does not disclose the distribution of performance gains under different problem scales or dimensions.
All Facts (3)
The paper proposes an algorithm named BO-pro-c-tr that uses a product-of-experts Gaussian process with uncertainty calibration (GP-pro-c) to重构 the paradigm of large-scale Bayesian optimization.
50%UnverifiedWhen combined with the GP-LCB acquisition function, BO-pro-c-tr reduces average simple regret by 2.6% and computational overhead by 34.9% relative to methods based on independent local GPs and trust regions.
50%UnverifiedThe paper does not disclose the distribution of performance gains under different problem scales or dimensions.
50%