Threshold Device

A threshold device is a type of processing unit in an artificial neural network that employs a step function (i.e., a Heaviside equation) as its activation function. That is, the unit has an associated threshold. If the net input to the unit is greater than the threshold, it turns on (eg., adopts an activity of 1). Otherwise, it turns off. Sometimes, "off" is represented by a value of 0; some researchers use -1 to represent an off state. The binary nature of the threshold device permits it to model the "all or none" law of neural activation (Levitan & Kaczmarek, 1991). Threshold devices were critical components of McCulloch-Pitts networks (McCulloch & Pitts, 1943) and served as the output units for perceptrons (Rosenblatt, 1962).

References:

1. Levitan, I. B., & Kaczmarek, L. K. (1991). The Neuron: Cell And Molecular Biology. New York: Oxford University Press.
2. McCulloch, W. S., & Pitts, W. (1943). A logical calculus of the ideas immanent in nervous activity. Bulletin of Mathematical Biophysics, 5, 115-133.
3. Rosenblatt, F. (1962). Principles Of Neurodynamics. Washington: Spartan Books.