Standard artificial neural networks are constructed from 'units' (roughly similar to neurons) that transmit their 'activity' (similar to membrane potentials or to mean firing rates) to other units via 'weight factors' (similar to synaptic coupling efficacies).