Neuropeptide-like operators in artificial neural networks

Recent biological evidence shows that some peptides may co-exist, and be co-released with classical neurotransmitters in synaptic terminals. These peptidergic components appear to modulate the signalling activity of their co-released neurotransmitters. This paper presents a simulation of a possible modulatory activity of neuropeptides in a backpropagation network, and contrasts it with the effects of random noise in the same network. This allows a description of the possible nature of peptide modulation as an effective learning enhancer, similar to noise in enhancement magnitude, but possibly different in mechanism.