A supervised learning algorithm of SNN is proposed by using spike sequences with complex spatio-temporal information. We explore an error back-propagation method of SNN based on gradient descent. The chain rule proved mathematically that it is sufficient to update the SNN’s synaptic weights by directly using an optimizer. Utilizing the TensorFlow framework, a bilayer supervised learning SNN is constructed from scratch. We take the lead in the application of SAR image classification and conduct experiments on the MSTAR dataset.
