Multibit neural inference in a N N -ary crossbar architecture
Anatole Moureaux, Anthony Lopes Temporao, Flavio Abreu Araujo · 2026
Read Multibit neural inference in a N N -ary crossbar architecture by Anatole Moureaux, Anthony Lopes Temporao, Flavio Abreu Araujo (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-2604-26979.txt.
Description
Moureaux and two colleagues present a tool for N-ary in-memory crossbar architectures that retrieve matrix-vector results with minimal assumptions. Such computing performs matrix-vector products in memory arrays, promising energy gains. XOR and MNIST tasks run on an array of 4 by 4 magnetic tunnel junctions, with MNIST accuracy near 94 percent versus a higher software baseline. PCA reduction sharply cuts operations and lifts the baseline, for only a modest crossbar score cost. Weight quantization is the top error source, and cell noise proves less harmful than errors.
Catalog record
| CREATOR | Anatole Moureaux, Anthony Lopes Temporao, Flavio Abreu Araujo |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 44959 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.26979