What is going to happen?
The
basic aim of this network is to learn and store a set of patterns so
it can recall them later - this is the network's memory. In this
case, the patterns are pictures of numbers (zero through nine)
represented as a set of on or off nodes (white
or black squares) in a two dimensional grid. When you hit
the "Start the Simulation" button, there will be a small delay
while the network connection strengths are adjusted to store/learn
each number's pattern. This is called `training'. Then,
after your network is 'trained,' we will distort pictures of numbers to
see how well the networks learned each number's patterns before it can't
tell what number it is anymore... but more on that later.
-
Why does it take so long?
- When you hit the "Start the Simulation" button, the
computer (that stores all this information) will start run Simon
Catterall's Hopfield Neural Network program. While you are staring at this
screen, the program reads all the data that is used to represent
the numbers 0-9 (approximately 2,560 weights). Then, each time that
you start from this web page, the Net will learn the pattern of
each number so that it can examine pictures, that you will help
create, and try to figure out what number is in that picture. So it
will take a couple of seconds for the computer to sift through the
data, and if more people are using this same page (across the campus) then
it will be a little slower! So Please, be patient.
-
What is this all about?
- If you are still wondering what exactly is going on, you may want
to go back and re-read (or read :) the parts of the Introdction to Neural Networks.
This will explain fully what happens with Neural Networks in Computer
Science.
Since this page will actually execute a complex program you will have to wait for it to finish. Be patient, you might have to wait 10-20 seconds