← 研踪 PaperTrailarXiv:2609.35628 · v2
Arbitrary-Accuracy Neural Approximation with Optimal Neuron Count and Optimal Bit Complexity
首次提交 2026-09-29 01:00
版本更新 2026-11-19 18:41
雷达观测 2026-11-20 19:57
摘要
We study the minimum number of hidden neurons required for arbitrary-accuracy approximation of continuous functions on and the associated encoding complexity. For , we construct two fixed, explicitly defined activation functions and, for each, a closed-form network with two hidden layers of widths and that achieves arbitrary accuracy in the uniform norm. We prove that is the exact minimum total number of hidden neurons among standard feedforward networks with locally integrable activations and affine outputs. We further give a simpler construction using a single elementary activation that combines the floor and exponential functions. This construction requires three hidden layers of widths , , and . If a skip connection is allowed, widths , , and suffice. For a bounded -Hölder class with , the networks with the digit-extraction activation and with the elementary activation require bits, matching the metric-entropy lower bound up to a logarithmic factor. The other network, which stores differences of neighboring values along a spanning tree, needs only bits and is optimal in both neuron count and bit complexity.
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v2 · 2026-11-19 18:41 · 最新观测版本
Arbitrary-Accuracy Neural Approximation with Optimal Neuron Count and Optimal Bit Complexity
We study the minimum number of hidden neurons required for arbitrary-accuracy approximation of continuous functions on and the associated encoding complexity. For , we construct two fixed, explicitly defined activation functions and, for each, a closed-form network with two hidden layers of widths and that achieves arbitrary accuracy in the uniform norm. We prove that is the exact minimum total number of hidden neurons among standard feedforward networks with locally integrable activations and affine outputs. We further give a simpler construction using a single elementary activation that combines the floor and exponential functions. This construction requires three hidden layers of widths , , and . If a skip connection is allowed, widths , , and suffice. For a bounded -Hölder class with , the networks with the digit-extraction activation and with the elementary activation require bits, matching the metric-entropy lower bound up to a logarithmic factor. The other network, which stores differences of neighboring values along a spanning tree, needs only bits and is optimal in both neuron count and bit complexity.
查看此版本解读 →查看此版本原文 ↗