SpikingGamma: Surrogate-Gradient Free and Temporally Precise Online Training of Spiking Neural Networks with Smoothed Delays

Roel Koopman, Sebastian Otte, Sander Bohté, [2pt] Machine Learning Group, CWI · 2026

Read SpikingGamma: Surrogate-Gradient Free and Temporally Precise Online Training of Spiking Neural Networks with Smoothed Delays by Roel Koopman, Sebastian Otte, Sander Bohté, [2pt] Machine Learning Group, CWI (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-2602-01978.txt.

Description

Roel Koopman and colleagues train spiking neural networks online with SpikingGamma, using smoothed delays and no surrogates. Neuromorphic hardware promises efficient low latency AI through sparse event computation, but fine temporal training remains hard, blocking hardware mapping. Current methods model spiking neurons as self recurrent units, training with BPTT or RTRL on surrogate gradients, scaling poorly with resolution. SpikingGamma combines neurons with recursive memory and sigma delta spike coding, with direct backpropagation. It learns fine temporal patterns with minimal spiking, scales feedforward networks to complex tasks at any resolution.

Catalog record

CREATORRoel Koopman, Sebastian Otte, Sander Bohté, [2pt] Machine Learning Group, CWI
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE48661
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602.01978

Browse the full library · About the Tower of Babel