Quantization of Spiking Neural Networks Beyond Accuracy Conference: ; ;

Evan Gibson Smith, Jacob Whitehill, Fatemeh Ganji · 2026

Read Quantization of Spiking Neural Networks Beyond Accuracy Conference: ; ; by Evan Gibson Smith, Jacob Whitehill, Fatemeh Ganji (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-2604-14487.txt.

Description

Smith, Whitehill, and Ganji argue spiking-network quantization should be judged beyond accuracy, on whether firing behavior survives. Method, clipping range, and bit-width shift firing distributions at equal accuracy, with effects standard metrics miss but deployment feels through sparsity, storage, and event-processing load. The team proposes Earth Mover's Distance as a firing-drift diagnostic, applied to weight and membrane quantization on SEW-ResNet models trained on CIFAR-10 and CIFAR-100. Uniform quantization drifts even as accuracy holds, while learned LQ-Net-style quantization stays near the full-precision baseline.

Catalog record

CREATOREvan Gibson Smith, Jacob Whitehill, Fatemeh Ganji
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE32125
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.14487

Browse the full library · About the Tower of Babel