Individual differences |
Methods | Statistics | Clinical | Educational | Industrial | Professional items | World psychology |
Bidirectional associative memory (BAM) is a type of recurrent neural network. BAM was introduced by Bart Kosko in 1988. There are two types of associative memory, auto-associative and hetero-associative. BAM is hetero-associative, meaning given a pattern it can return another pattern which is potentially of a different size. It is similar to the Hopfield network in that they are both forms of associative memory. However, Hopfield nets return patterns of the same size.
Topology[edit | edit source]
A BAM contains two layers of neurons, which we shall denote X and Y. Layers X and Y are fully connected to each other. Once the weights have been established, input into layer X presents the pattern in layer Y, and vice versa.
Procedure[edit | edit source]
Learning[edit | edit source]
Imagine we wish to store two associations, A1:B1 and A2:B2.
- A1 = (1, 0, 1, 0, 1, 0), B1 = (1, 1, 0, 0)
- A2 = (1, 1, 1, 0, 0, 0), B2 = (1, 0, 1, 0)
These are then transformed into the bipolar forms:
- X1 = (1, -1, 1, -1, 1, -1), Y1 = (1, 1, -1, -1)
- X2 = (1, 1, 1, -1, -1, -1), Y2 = (1, -1, 1, -1)
From there, we calculate where denotes the transpose. So,
Recall[edit | edit source]
To retrieve the association A1, we multiply it by M to get (4, 2, -2, -4), which, when run through a threshold, yields (1, 1, 0, 0), which is B1. To find the reverse association, multiply this by the transpose of M.
See also[edit | edit source]
References[edit | edit source]
[edit | edit source]
- Bidirectional Associative Memory - online demo and perl source code
- Bidirectional associative memories - ACM Portal Reference
|This page uses Creative Commons Licensed content from Wikipedia (view authors).|