Feedforward
Associative Memory Networks
As it is shown in the next figure,
a basic association network has one layer of input neurons and one layer
of output neurons. An input pattern x is represented as a point in the
N-dimensional real or binary vector space, denoted as 
or 
, respectively.
Linear Associative Memory (LAM)
An associative memory network is matematically a mapping from an input
space. Associative memory networks can be applied to either auto-association
or hetero-association applications. In the auto-association application,
the dimension of the input space is equal to that of the output space.
On the other hand, in the hetero-association application, the dimension
of the input space and the output space are in general different. The input
and outputs values can be real or binary.
A LAM is a single-layer feedforward
network. The LAM is derived from a set
of input/output pattern pairs 
. Here the input is
, and the output is 
for m=1, 2,...., M, where 
denotes
the transpose of a vector or matrix.The objective of LAM is to recover
the output pattern based on the full or partial information of the input
pattern. We will review two cases: Continuous-Valued
Input Patterns and Binary-Valued Input Patterns
.
Continuous-Valued Input Patterns
Contents
Artificial Neural Networks
About this Tutorial