Work

GABE: Groupwise Affine Basis Encoding

Year: 2026

Type: article

Abstract: Introduce GABE (Groupwise Affine Basis Encoding) - a decomposition method that represents neural network weights as an addressable memory system. For any group of similar layers, we extract three interpretable components: a shared mean weight (long-term memory), a low-rank basis of variations (address space), and per-layer coefficients (pointers).

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.18488810

Author: Dmitry Feklin

Language: English

Topic: Explainable Artificial Intelligence (XAI)

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

OpenAlex ID: W7127623954

FWCI:

Cites:

Cited by:

Related to: