Scale and translation invariant learning of spatio-temporal patterns using longest common subsequences and spiking neural networks.