What Is a Hopfield Neural Network?

Hopfield neural network is a recursive neural network invented by John Hopfield in 1982.

Hopfield neural network is a

The units of the Hopfield network are binary, that is, these units can only accept two different values, and the value depends on whether the size of the input reaches the threshold. Hopfield networks usually accept values of -1 or 1, and can also be 0 or 1. Input is by

Associative memory is an important application area of discrete Hopfield networks. To achieve associative memory, the feedback network must have two basic conditions:

The network can converge to a stable equilibrium state, and use it as the memory information of the sample;

It has the ability to recall, and can recall the complete memory information that it belongs to from a certain incomplete information. The process of implementing associative memory in discrete Hopfield networks is divided into two phases: the learning and memory phase and the associative memory phase. In the learning and memory phase, the designer determines a suitable set of weights through a certain design method to make the network memory expect a stable equilibrium point. Lenovo recall stage is the working process of the network.

Discrete Hopfield networks used in associative memory have two outstanding characteristics: that memory is distributed, while associativity is dynamic. The limitations of discrete Hopfield networks are mainly reflected in the following points:

the limitation of memory capacity;

Association and memory of pseudo stable points;

When the memory samples are close, the network cannot always recall the correct memory. In addition, the balance and stability point of the network cannot be set arbitrarily, and there is no universal way to know the balance and stability point in advance.

• Boltzmann Machine-Like a Hopfield Network, Annealed Gibbs Sampling Can Be Used Instead of Gradient Descent

• Ising model

• Herb theory