Binarized neural networks converge toward algorithmic simplicity: empirical support for the learning-as-compression hypothesis

Understanding and controlling the complexity of neural networks is a central challenge in machine learning, with implications for generalization, optimization, and model capacity. While most approaches rely on entropy-based loss functions and statistical metrics, these measures often fail to capture...

Full description

Bibliographic Details

Published in:	Frontiers in Computational Neuroscience
Main Authors:	Eduardo Y. Sakabe, Felipe S. Abrahão, Alexandre Simões, Esther Colombini, Paula Costa, Ricardo Gudwin, Hector Zenil
Format:	Article
Language:	English
Published:	Frontiers Media S.A. 2026-05-01
Subjects:	algorithmic information theory binarized neural networks Block Decomposition Method learning as compression Shannon entropy
Online Access:	https://www.frontiersin.org/articles/10.3389/fncom.2026.1791546/full