A Novel Schur-Decomposition-Based Weight Projection Method for Stable State-Space Neural-Network Architectures

Sergio Vanegas, Lasse Lensu, Fredy Ruiz · 2026

Read A Novel Schur-Decomposition-Based Weight Projection Method for Stable State-Space Neural-Network Architectures by Sergio Vanegas, Lasse Lensu, Fredy Ruiz (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-2605-14489.txt.

Description

Vanegas, Lensu and Ruiz present a stability-ensuring, backpropagation-compatible Schur projection scheme for linear discrete-time state-space layers. The methods project the quasi-triangular factor of the real Schur decomposition onto its nearest stable peer. Tests on synthetic linear systems show accuracy and convergence matching state-of-the-art methods, despite higher compute cost. The lower weight count speeds convergence during training without losing accuracy.

Catalog record

CREATORSergio Vanegas, Lasse Lensu, Fredy Ruiz
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE80614
LICENSEhttp://creativecommons.org/licenses/by-sa/4.0/
RIGHTSCC BY SA
SOURCEarXiv

Source: https://arxiv.org/abs/2605.14489

Browse the full library · About the Tower of Babel