A neuron is a single cell in the brain or item in an artificial neural network. each neuron is very simple, but the overall behaviour of the netwrok can be very rich. Each neuron tpyically has a number of inputs form other neurons and can 'fire' if the tital amount of input activation is high. This may then cause other neurons to fire. The simplest form of artificial neoruon is a perceptron, which simply takes a weighted sum of its inputs and then first if these exceed a hard {[threshold}}. The neurons used in current neural networks typically have a softer threshold function to enable learning through backpropagation. Real neurons may have complex time-dependant behaviour, for example being more or less likely to fire again if they have recently fired, and fire in a bursty and asynchonous manner. Some {[artificial neural networks}} try to emulate aspects of this; for eample {{long-term short-term memory networks}] and spiking neural networks.
Used in Chap. 6: pages 84, 90, 92; Chap. 9: page 137
A single perceptron.