constructive-approximation

The paper introduces Variation Brownian Kernel Ladders (VBKL), a path-atomic function-space framework for recursive dictionary construction, with analysis on Hölder regularity, compactness, and generalization bounds in statistical learning theory.