Practical Bayesian Inference for Speech SNNs: Uncertainty and Loss-Landscape Smoothing

Yesmine Abdennadher, Philip N. Garner · 2026

Read Practical Bayesian Inference for Speech SNNs: Uncertainty and Loss-Landscape Smoothing by Yesmine Abdennadher, Philip N. Garner (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-2604-08624.txt.

Description

Abdennadher and Garner bring Bayesian learning to spiking neural networks for speech, smoothing their jagged predictive landscape. Threshold based spike generation makes predictions angular and irregular, but a Bayesian treatment of weights, including the efficient IVON variational method, yields smoother landscapes. Tests on Heidelberg Digits and Speech Commands improve negative log likelihood and Brier score, with one dimensional weight slices confirming the smoother landscape.

Catalog record

CREATORYesmine Abdennadher, Philip N. Garner
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE27546
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.08624

Browse the full library · About the Tower of Babel