MAP-SNN training method

Multiplicity, adaptability and plasticity in bio-plausible SNN training

A training method addressing the time-discretization problem in spiking neural networks while improving bio-plausibility.

  • An algorithm investigating the discretization problem in time-iteration, demonstrating robustness under varying iterative step lengths.
  • A spike-frequency-adaptation mode as a multiple-spike-pattern implementation, for efficient spike inference.
  • A State-Free Synaptic Response Model (SFSRM) enhancing temporal expressiveness, compatible with deep learning frameworks.

Published in Frontiers in Neuroscience, 2022.