Tag
This paper introduces ALAS, a flexible Gaussian Process kernel family that learns the stability parameter from data to adapt smoothness, capturing both smooth trends and sharp irregularities, with a separable variant for higher dimensions and theoretical guarantees on information gain.