FTBC: Forward Temporal Bias Correction for Optimizing ANN-SNN Conversion

arXiv:2403.18388v1 Announce Type: cross
Abstract: Spiking Neural Networks (SNNs) offer a promising avenue for energy-efficient computing compared with Artificial Neural Networks (ANNs), closely mirroring biological neural processes. However, this potential comes with inherent challenges in directly training SNNs through spatio-temporal backpropagation -- stemming from the temporal dynamics of spiking neurons and their discrete signal processing -- which necessitates alternative ways of training, most notably through ANN-SNN conversion. In this work, we introduce a lightweight Forward Temporal Bias Correction (FTBC) technique, …

abstract ann anns artificial artificial neural networks arxiv backpropagation bias challenges computing conversion cs.ai cs.cv dynamics energy however networks neural networks neurons processes snn spiking neural networks stemming temporal through training type