Node Model

A neural network node is an object that is responsible for carrying out very basic logical tasks by manipulating small amounts of data in very unique ways. When many nodes are connected to one another in a certain way, the network they create is capable of processing large amounts of data in much more complicated ways.

Many different network models can be created for the specific way a node functions. The efficiency of the model is highly dependent upon the rules and architecture of the network itself. Coming up with the best model is the challenge for this project.

This article will describe the minimum requirements a node must include in order to be of any practical use in a network.

nodeMinimum Requirements

Each node must have the following.

  • One or more input sockets
  • One or more output paths
  • At least one internal logic type identifier

A node’s responsibility is to interpret data and react based on its internal logic type identifier.

Figure 1 is an example of a node with two input sockets and one output path. Using what we know about the logic type identifier (described in Representing Logic), we can predict how this node will behave in a binary environment. (active or inactive)

  • When both A and B are active, the node will send out an active signal to its output path.
  • When A is active and B is not active, the node will send out an active signal to its output path.
  • When A is not active and B is active, the node will not send out a signal.
  • When neither A nor B are active, the node will send out an active signal to its output path.

All this can be determined just by looking at the node’s logic type identifier.

node2Multiple Logic Type Identifiers

It’s possible to associate one logic type identifier with each output path allowing the node to output different results using the same input. This is helpful when implementing more advanced algorithms in your network model.