Some stuff from Iris Oved's course.
Course page: http://www.u.arizona.edu/~zemel/Courses/psyNN.html
From Assignment 1: Faces
From Assignment 2:
simpleHopfield.ps
And here's what I figured out from studying the MatLab source
for "hopf". It takes three strings (names, phone numbers) and
encodes them as binary (using modified ascii codes) then
changes the binary 0 & 1 encoding to -1 & 1. These become the
initial patterns in a Hopfield setup as described in my notes.
Then a test pattern is iterated through the system by picking
random entries and updating them. There is a maximum iteration
count, but the process is aborted early if the energy remains
constant for too long.