C. Rizzo, J. Solanki, S. S> Mondal, N. Cady, C. D. Schuman, J. S. Plank, G. Rose and H. Das

August, 2026

ICONS: International Conference on Neuromorphic Systems

https://dl.acm.org/doi/10.1145/3822454.3822483

PDF not available yet, or is only available from the conference/journal publisher.

Abstract

Neuromorphic computing has emerged as a promising paradigm for ultra-low-power, high-density edge intelligence, where memristive synapses enable compact multi-bit weight storage. -based devices offer high integration density but suffer from intrinsic variability, particularly in the high-resistance state (HRS), leading to potential weight overlap and accuracy degradation. In this work, we present a comprehensive evaluation of synaptic noise in neuroprocessors integrating memristor-based synapses with CMOS neuron circuits. Experimentally characterized devices in 65 nm CMOS are analyzed using 3σ resistance distributions, and the extracted noise profiles are incorporated into two neuromorphic architectures: RISP (integrate-and-fire) and RAVENS (leaky integrate-and-fire with refractory behavior). We further study encoding schemes under noise across classification, control, and regression tasks. Results show strong resilience to synaptic variability, with limited performance degradation under realistic conditions. Additionally, measured inference power averages 3.85 μW in LRS and 0.79 μW in HRS, enabling nearly power savings with negligible performance loss, highlighting the potential for variability-tolerant, energy-efficient edge AI systems.

Citation Information

Text

author      C. Rizzo and J. Solanki and S. S> Mondal and N. Cady and C. D. Schuman and J. S. Plank and G. Rose and H. Das
title       Neuroprocessor Noise Resilience Using {HfO2}-Based Synapses
booktitle   International Conference on Neuromorphic Systems (ICONS)
year        2026
url         https://dl.acm.org/doi/10.1145/3822454.3822483
doi         10.1145/3822454.3822483
pages       185-192

Bibtex

@INPROCEEDINGS{rsm:26:nnr,
    author = "C. Rizzo and J. Solanki and S. S> Mondal and N. Cady and C. D. Schuman and J. S. Plank and G. Rose and H. Das",
    title = "Neuroprocessor Noise Resilience Using {HfO2}-Based Synapses",
    booktitle = "International Conference on Neuromorphic Systems (ICONS)",
    year = "2026",
    url = "https://dl.acm.org/doi/10.1145/3822454.3822483",
    doi = "10.1145/3822454.3822483",
    pages = "185-192"
}