What question did this study set out to answer?

The aim is to improve neural sequence models by integrating hyperbolic geometry and dynamical systems.

January 24, 2026Open Access

Lorentz-Manifold Transformers: A Geometric–Dynamical Framework for Hierarchical Representation Learning

Puntos clave

The aim is to improve neural sequence models by integrating hyperbolic geometry and dynamical systems.
Introduced the Lorentz-Manifold Transformer (LMT) model.
Integrated hyperbolic geometry with oscillatory dynamics using Hyperbolic Artificial Kuramoto Oscillatory Neurons.
Developed mathematical guarantees for topological preservation.
Demonstrated exponential advantages in manifold capacity bounds.
Identified geometric frustration as a measure for representational misalignment.
Utilized Gromov-Wasserstein structural risk as a key evaluation metric.

Resumen

Neural sequence models face a fundamental tension between representational capacityand geometric constraints. We present the Lorentz-Manifold Transformer (LMT), integrat-ing hyperbolic geometry (Lorentz model) with oscillatory dynamics (Hyperbolic ArtificialKuramoto Oscillatory Neurons, H-AKOrN) to address the Geometric Capacity Bottle-neck. The LMT establishes mathematical guarantees for topological preservation through:(1) manifold capacity bounds proving exponential advantage (αcH /αcR = Ω(er )), (2) geomet-ric frustration as a proxy for representational misalignment, and (3) Gromov-Wassersteinstructural risk. Computational complexity (

Leer artículo completoexternamente

Preguntar a la IA

Me gusta

Guardar

Ver artículo completo

Cite This Study

E. G. Reis (Mon,) studied this question.

synapsesocial.com/papers/697460e9bb9d90c67120ac0a https://doi.org/https://doi.org/10.5281/zenodo.18334083

Preguntar a la IA

Me gusta

Guardar

Ver artículo completo