Explain Hopfield networks.

      ·         Hopfield introduced a neural network as a theory of memory.

      

      ·         Hopfield network is a collection of N neurons, which are fully connected.

      ·         Its features are – (DD-MM-PP-F)

1.      Distributed representation.

2.     Distributed asynchronous control.

3.     Memory is content addressable.

4.     Memory is stored as a pattern of activation across a set of processing elements.

5.     Each processing element makes decisions based only on its own local situation.

6.     A no. of patterns can be stored in the network. To retrieve a pattern, a specific portion of the pattern is specified, and the network automatically finds the closest match.

Fault tolerance: If a few processing elements misbehave or fail completely, the network will still function properly.

A simple Hopfield network is shown in the figure below-

·         The processing elements in a Hopfield network are always in one of two states – Active or Inactive.

       ·         Units are connected to each other with weighted symmetric connection.

      ·         A +ve weighted connection indicates that the 2 units tend to activate each other, while a –ve weighted connection allows an active unit to deactivate a neighboring unit.

Usability of Hopfield networks:

The network can store patterns as its memory by setting only part of the nodes. It can do so using its self-learning capability, using the weight matrix.

·         It uses Content Addressable Memory (CAM) to store and retrieve the patterns, from the weight matrix.

·       The networks can also be used for auto associations.