Spiking neural networks (SNNs) are considered as the 3rd generation of artificial neural networks, which closely mimic biological brains. However, such networks remain hard to train. We have been developing new computational methods to better train SNNs using Bayesian inference.