Advancing Constrained Monotonic Neural Networks: Achieving Universal Approximation Beyond Bounded Activations

Davide Sartor, Alberto Sinigaglia, Gian Antonio Susto · 2025

Read Advancing Constrained Monotonic Neural Networks: Achieving Universal Approximation Beyond Bounded Activations by Davide Sartor, Alberto Sinigaglia, Gian Antonio Susto (2025) 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-2505-02537.txt.

Description

Davide Sartor, Alberto Sinigaglia, and Gian Antonio Susto widen the theory of one-way neural networks in this 2025 paper. Such networks must never decrease their output when inputs rise, a rule enforced by weight limits and capped activation curves. Those caps make training awkward.

The team proves that networks with caps that flatten on alternating sides can still mimic any one-way function. They also show the cap side and the weight sign must match in a set way. A variant lets the network tune its own caps to the weight signs, easing setup and steadying training. Tests back the theory against classic designs.

Catalog record

CREATORDavide Sartor, Alberto Sinigaglia, Gian Antonio Susto
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE76 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-02537

Browse the full library · About the Tower of Babel