Pith. sign in

REVIEW

Toward High Performance, Programmable Extreme-Edge Intelligence for Neuromorphic Vision Sensors utilizing Magnetic Domain Wall Motion-based MTJ

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2402.15121 v1 pith:ODVOGTEE submitted 2024-02-23 cs.AR cs.ETeess.IV

Toward High Performance, Programmable Extreme-Edge Intelligence for Neuromorphic Vision Sensors utilizing Magnetic Domain Wall Motion-based MTJ

classification cs.AR cs.ETeess.IV
keywords highenergymagneticneuromorphicvisionaccuracyconstraintsconsumption
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

The desire to empower resource-limited edge devices with computer vision (CV) must overcome the high energy consumption of collecting and processing vast sensory data. To address the challenge, this work proposes an energy-efficient non-von-Neumann in-pixel processing solution for neuromorphic vision sensors employing emerging (X) magnetic domain wall magnetic tunnel junction (MDWMTJ) for the first time, in conjunction with CMOS-based neuromorphic pixels. Our hybrid CMOS+X approach performs in-situ massively parallel asynchronous analog convolution, exhibiting low power consumption and high accuracy across various CV applications by leveraging the non-volatility and programmability of the MDWMTJ. Moreover, our developed device-circuit-algorithm co-design framework captures device constraints (low tunnel-magnetoresistance, low dynamic range) and circuit constraints (non-linearity, process variation, area consideration) based on monte-carlo simulations and device parameters utilizing GF22nm FD-SOI technology. Our experimental results suggest we can achieve an average of 45.3% reduction in backend-processor energy, maintaining similar front-end energy compared to the state-of-the-art and high accuracy of 79.17% and 95.99% on the DVS-CIFAR10 and IBM DVS128-Gesture datasets, respectively.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.