|
Neural
Networks |
What is Neural Computing?
The study of massively parallel
network of simple processors that have a natural tendency for storing
experiential knowledge. Knowledge is acquired through training rather than
programming and is stored in the connection strengths. A neural network
consists of many simple processors which work together to solve a
particular problem.
The processors are linked via an arrangement
of weighted connections. These weights are altered during training and
thus a model of the problem domain is stored in a distributed manner
across the networks weights.
Biological Neural Network
An
Artificial Neuron Model
- The input pattern [X1..Xn] consists of n attributes which describe
an object or event from domain of interest (e.g. height, weight, age
etc. to describe people)
- Each attribute value is passed along the connection and multiplied
by the weight W on that connection.
- The sum of these weighted attribute values forms the neurons
activation net.
- The neuron output would be found by passing net through a threshold
function
Further Reading
[1] An Introduction to Neural Computing (2nd Edition), I Aleksander and
H Morton, Thompson, 1995
[2] Artificial Neural Networks, D Patterson, Prentice Hall, 1996
[3] Fundamentals of Neural Networks, L Fausset, Prentic Hall, 1994 |
