Diagnostic toolkit for GP boundary bias in Bayesian optimisation
Provides PyTorch/NumPy code to reproduce the boundary‑variance bias paper, including kernels, GP posterior, acquisition functions, and diagnostic plots. Researchers and practitioners can run experiments, generate figures, and explore how variance‑driven acquisitions mis‑place points near domain edges. It offers a focused, reproducible study not found in standard BO libraries.
View on GitHub →mariabankestad/gp-boundary-bias