Foundations Of Cognitive Science

Hidden Unit

A hidden unit refers to the components comprising the layers of processors between input and output units in a connectionist system.  The hidden units add immense, and necessary, power to connectionism, in that their advent was a precursor to the generalized delta rule, and allowing them to be equal in power to universal Turing machines by giving them arbitrary pattern classifier and universal function approximation powers.  These units were important in the history of connectionism as well.  While old connectionism had such units in some of their networks, it had no method to train them.  New connectionism arose when researchers discovered learning rules that could modify all of the connections in multilayer networks (Ackley et al, 1985; Rumelhart et al., 1986).


  1. Ackley, D. H., Hinton, G. E., & Sejnowski, T. J. (1985). A learning algorithm for Boltzman machines. Cognitive Science, 9, 147-169.
  2. Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986). Learning representations by back-propagating errors. Nature, 323, 533-536.
    (Added November 2009)