Multibit neural inference in a N-ary crossbar architecture

TL;DR

This paper introduces a simulation framework for N-ary crossbar architectures in in-memory computing, demonstrating neural inference with magnetic tunnel junctions and analyzing error sources to optimize performance.

Contribution

It presents a novel simulation approach for N-ary crossbar arrays and investigates error impacts, including quantization and noise, to improve neural inference accuracy.

Findings

Achieved 94.48% MNIST accuracy with simulated 4-state MTJ crossbar.

Identified weight quantization as the main error source affecting inference.

Optimal number of states per cell balances quantization error and resistance resolution.

Abstract

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.