The neurons in an Artificial Neural Network (ANN) are typically simple compared to real biological neurons. Artificial neurons don't require fuel in the form of a blood supply. They don't interact with astrocytes when signalling neighboring neurons. To find out how important these differences might be, researchers Xi Shen and Philippe de Wilde have developed a more realistic neural network simulation that includes neurons, glial cells, capillaries with variable dialation to supply blood, bidirection neuron-glial cell signalling, and astrocyte interactions with neuron excitation. They believe their new model shows some interesting emergent behaviours as well as evidence of an attractor instead of random behavior exhibited by similarly structured neuron-only networks. The details can be found in their paper, Long-term Neuronal Behavior Caused by Two Synaptic Modification Mechanisms (PDF format).