← back to paper
arxiv: 2602.07618 · 2 revisions
Neural Networks With Dense Weights Are Not Universal Approximators