sakuya/node_modules/neataptic/mkdocs/templates/docs/builtins/random.md

description: How to use the Random model network in Neataptic authors: Thomas Wagenaar keywords: recurrent, feed-forward, gates, neural-network, random, architecture

A random network is similar to a liquid network. This network will start of with a given pool of nodes, and will then create random connections between them. This network is really only useful for the initialization of the population for a genetic algorithm.

new architect.Random(input_size, hidden_size, output_size, options);

• input_size : amount of input nodes
• hidden_size : amount of nodes inbetween input and output
• output_size : amount of output nodes

Options:

• connections : amount of connections (default is 2 * hidden_size, should always be bigger than hidden_size!)
• backconnections : amount of recurrent connections (default is 0)
• selfconnections : amount of selfconnections (default is 0)
• gates : amount of gates (default is 0)

For example:

var network = architect.Random(1, 20, 2, {
  connections: 40,
  gates: 4,
  selfconnections: 4
});

drawGraph(network.graph(1000, 800), '.svg');

will produce: