Graded Neural Networks

Tanush Shaska

I-AIMS paper sh-89 · Published in Int. J. Data Sci. Math. Sci. 3 (2025), no. 2, 87–116 · 2025

We introduce a rigorous framework for Graded Neural Networks, a new class of architectures built on coordinate-wise graded vector spaces 𝒱ⁿ_q. Using an algebraic scalar action λ ⋆ x = (λ^(qᵢ) xᵢ) defined by a grading tuple q = (q₀, …, qₙ₋₁), we construct grade-sensitive neurons, activations, and loss functions that embed hierarchical feature structure directly into the network’s architecture. This grading endows GNNs with enhanced expressivity and interpretability, extending classical neural networks as a special case. We develop both the algebraic theory and computational implementation of GNNs, addressing challenges such as numerical instability and optimization with anisotropic scaling. Theoretical results establish universal approximation for graded-homogeneous functions, along with convergence rates in graded Sobolev and Besov spaces. We also show that GNNs achieve lower complexity for approximating structured functions compared to standard networks. Applications span hierarchical data modeling, quantum systems, and photonic hardware, where grades correspond to physical parameters. This work provides a principled foundation for incorporating grading into neural computation, unifying algebraic structure with learning, and opening new directions in both theory and practice.

MSC classes: Primary 68T07; Secondary 16W50, 13A02

PDF

References

1. T. Shaska, Artificial neural networks on graded vector spaces, Contemporary Mathematics (2025).
2. Elira Shaska and Tanush Shaska, Machine learning for moduli space of genus two curves and an application to isogeny-based cryptography, J. Algebraic Combin. 61 (2025), no. 2, Paper No. 23, 35.
3. N. Bourbaki, Algebra I, Springer, 1974. Chapter 3.
4. Steven Roman, Advanced linear algebra, Third, Graduate Texts in Mathematics, vol. 135, Springer, New York, 2008.
5. J.-L. Koszul, Graded manifolds and graded Lie algebras, Proceedings of the international meeting on geometry and physics (Florence, 1982), 1983, pp. 71–84.
6. Songpon Sriwongsa and Keng Wiboonton, The triangle inequality for graded real vector spaces, Math. Inequal. Appl. 23 (2020), no. 1, 351–355.
7. Martin Moskowitz, The triangle inequality for graded real vector spaces of length 3 and 4, Math. Inequal. Appl. 17 (2014), no. 3, 1027–1030.
8. ———, An extension of Minkowski’s theorem to simply connected 2-step nilpotent groups, Port. Math. 67 (2010), no. 4, 541–546.
9. Sajad Salami and Tony Shaska, Local and global heights on weighted projective varieties, Houston J. Math. 49 (2023), no. 3, 603–636.
10. Stephen Boyd and Lieven Vandenberghe, Convex optimization, Cambridge University Press, 2004.
11. Sajad Salami and Tony Shaska, Vojta’s conjecture on weighted projective varieties, Eur. J. Math. 11 (2025), no. 1, Paper No. 12, 33.
12. Ian Goodfellow, Yoshua Bengio, and Aaron Courville, Deep learning, MIT Press, 2016.
13. Yikun Nie, Bo Yang, Dongliang Wang, Ting Wang, Jiawei Wang, Zihao Wang, and Chaoran Huang, Integrated laser graded neuron enabling high-speed reservoir computing without a feedback loop, Optica 11 (2024Dec), no. 12, 1690–1699.
14. Kurt Hornik, Approximation capabilities of multilayer feedforward networks, Neural Networks 4 (1991), no. 2, 251–257.
15. Dmitry Yarotsky, Error bounds for approximations with deep relu networks, Neural Networks 94 (2017), 103–114.
16. Ronald A. DeVore, Nonlinear approximation, Acta Numerica 7 (1998), 51–150.

All preprints