Progressive Approximation in Deep Residual Networks: Theory and Validation

Wei Wang, Xiao-Yong Wei, Qing Li · 2026

Read Progressive Approximation in Deep Residual Networks: Theory and Validation by Wei Wang, Xiao-Yong Wei, Qing Li (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-24154.txt.

Description

Wang, Wei, and Li reframe residual networks as layer-wise approximation, tracking how error evolves from input to target. The universal approximation theorem promises approximation but says nothing about how layers share the work. The authors prove progressive trajectories exist on which error falls steadily with depth, then propose layer-wise progressive training, a rule aligning each layer with its residual target. The behavior appears in residual feedforward nets, ResNets, and Transformers, and one trained network yields useful predictions at every depth, enabling shallow use without retraining.

Catalog record

CREATORWei Wang, Xiao-Yong Wei, Qing Li
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE86330
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.24154

Browse the full library · About the Tower of Babel