Tensor‑Augmented Convolutional Neural Networks: Enhancing Expressivity with Generic Tensor Kernels

Chia-Wei Hsing, Wei-Lin Tu · 2026

Read Tensor‑Augmented Convolutional Neural Networks: Enhancing Expressivity with Generic Tensor Kernels by Chia-Wei Hsing, Wei-Lin Tu (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-2604-08072.txt.

Description

Hsing and Tu propose tensor-augmented CNNs, a physics-guided shallow model for richer feature learning. CNNs extract local features in layers, but capturing complex correlations usually demands deep, compute-heavy, hard-to-interpret architectures. Their tensor-augmented CNN replaces standard kernels with generic tensors to boost representational power. The reason is physical: an order-N tensor holds an arbitrary quantum superposition state in a Hilbert space of dimension d to the power N. Each layer's output becomes a multilinear form capturing high-order feature correlations.

Catalog record

CREATORChia-Wei Hsing, Wei-Lin Tu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE35547
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.08072

Browse the full library · About the Tower of Babel