The Perceptron Algorithm
Rune Sundgot, Tor H. Gulliksen · NORA - Norwegian Open Research Archives · 1995
The perceptron algorithm is an iterative method for solving a system of linear inequalities.We study the performance of this algorithm when used as a "learning" algorithm in a type of artificial neural networks called perceptrons.We show that the number of iterations needed to reach a solution can grow hyperexponentially in the dimensionality of the problem.We also extend an earlier result concerning the behaviour of this algorithm in the case where no solution exists.