ABSTRACT
As we have stressed in earlier chapters, neural networks are large interacting systems of simple units, like the physical systems we study in statistical mechanics. The formal methods and concepts of statistical physics are therefore natural tools to use for neural networks. In this chapter we illustrate the use of such methods in two different problems that we encountered earlier in the book: the recall of stored patterns in the Hopfield associative memory network, and the capacity of a simple perceptron.
