adaptive-basis

This paper introduces Data-Driven Variational Basis Learning (DVBL), a non-neural framework that learns basis functions directly from data through variational optimization, offering interpretability and mathematical transparency compared to neural networks.