Concept独立局部高斯过程 / Independent Local Gaussian Process
GP-ind
A method for large-scale Bayesian optimization that partitions the search space into multiple independent subregions, trains a local Gaussian process in each subregion, and optimizes with a trust-region mechanism. The approach is computationally efficient but suffers from discontinuities at region boundaries and loss of long-range information between subregions.
Timeline (last 90 days)
Nov 3
The GP-ind (independent local Gaussian process) with trust region approach in large-scale Bayesian optimization produces discontinuous region boundaries, loses long-range correlations, and causes a sharp increase in computational cost.
Unverified50%