Salvato in:
Dettagli Bibliografici
Autori principali: Moureaux, Anatole, Temporao, Anthony Lopes, Araujo, Flavio Abreu
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:https://arxiv.org/abs/2604.26979
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866914517474082816
author Moureaux, Anatole
Temporao, Anthony Lopes
Araujo, Flavio Abreu
author_facet Moureaux, Anatole
Temporao, Anthony Lopes
Araujo, Flavio Abreu
contents In-memory computing (IMC) enables energy-efficient neural network inference by computing analog matrix-vector multiplications (MVM) in memory crossbar arrays. In this work we present a simulation framework for N-ary crossbar architectures that retrieves MVM results with minimal implementation assumptions. The XOR and MNIST classification tasks were successfully inferred using a simulated crossbar array of (4x4) 4-states magnetic tunnel junctions (MTJ). MNIST accuracy reached 94.48% (vs. 97.56% software baseline). The software-hardware performance gap was further reduced using PCA dimensionality reduction. We identified weight quantization as the primary error source, and studied its impact alongside systematic nonidealities and random noise. We find that cell-specific random noise is less detrimental than systematic errors due to averaging across the array. Finally, we demonstrate an optimal number of states per cell that balances quantization error against resistance state resolution to minimize total MVM error.
format Preprint
id arxiv_https___arxiv_org_abs_2604_26979
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Multibit neural inference in a N-ary crossbar architecture
Moureaux, Anatole
Temporao, Anthony Lopes
Araujo, Flavio Abreu
Hardware Architecture
Artificial Intelligence
Emerging Technologies
In-memory computing (IMC) enables energy-efficient neural network inference by computing analog matrix-vector multiplications (MVM) in memory crossbar arrays. In this work we present a simulation framework for N-ary crossbar architectures that retrieves MVM results with minimal implementation assumptions. The XOR and MNIST classification tasks were successfully inferred using a simulated crossbar array of (4x4) 4-states magnetic tunnel junctions (MTJ). MNIST accuracy reached 94.48% (vs. 97.56% software baseline). The software-hardware performance gap was further reduced using PCA dimensionality reduction. We identified weight quantization as the primary error source, and studied its impact alongside systematic nonidealities and random noise. We find that cell-specific random noise is less detrimental than systematic errors due to averaging across the array. Finally, we demonstrate an optimal number of states per cell that balances quantization error against resistance state resolution to minimize total MVM error.
title Multibit neural inference in a N-ary crossbar architecture
topic Hardware Architecture
Artificial Intelligence
Emerging Technologies
url https://arxiv.org/abs/2604.26979