A type of neural network
that, unlike the Perceptron network
s, Adaline network
s, etc, is flat (single-layer) and interconnected
. This single layer acts as the input and output layer. Of course, this leads to the same number of outputs as inputs. This property, and the fact that an Hopfield net in a discrete
system will always isolate to either a single solution or a cyclic solution (if run interatively
), makes this network fairly useful for image/pattern recognition
Unlike Madaline networks, Hopfield nets have a lyapunov exponent of 0, noting that they do not exhibit chaos.