Product Interaction: An Algebraic Formalism for Deep Learning Architectures
Haonan Dong, Chun-Wun Cheng, Angelica I. Aviles-Rivero · 2026
Read Product Interaction: An Algebraic Formalism for Deep Learning Architectures by Haonan Dong, Chun-Wun Cheng, Angelica I. Aviles-Rivero (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-2602-02573.txt.
Description
Haonan Dong and 2 co-authors rebuild neural network design on algebra, treating multiplication as the basic act, in a 2026 arXiv paper. Convolutional, attention, state-space, and equivariant networks are usually built as separate crafts despite sharing recurring principles. Their scheme, product interactions, treats data as algebraic elements and builds networks by stacking multiplication operators.
How inputs act on themselves sets the order. Within the scheme, convolutional and equivariant networks emerge as symmetry-bound first-order cases. Attention arises as a third-order interaction from reusing the same operator. The Mamba design is recovered by raising a state-space model's interaction order from one to three. Such rules produce familiar structures: translation symmetry yields convolution, and rotational fit yields the Clebsch-Gordan product. A new simple layer built from these ideas beat Mamba on CIFAR-10 in both accuracy and throughput.
Catalog record
| CREATOR | Haonan Dong, Chun-Wun Cheng, Angelica I. Aviles-Rivero |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 89 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-02573