• Installation
• Usage
  • 1. Initialize a new population
  • 2. Evolve the population
• Examples
  • Xor
• License

Install using npm:

npm install neuroevolution

or Yarn yarn:

yarn add neuroevolution

The first step should be to initialize a new population.

A population accepts 3 main parameters:

• populationSize Total size of the genomes population
• nbInput Number of input nodes
• nbOutput Number of output node

const { Population } = require('neuroevolution')
const population = new Population(100, 2, 4)

This will create a population of 100 neural networks with 2 inputs and 4 outputs.

You can start evolving the population using evolve with 2 parameters :

• iterations How many iterations to evolve
• fitnessFunction You fitness function that has access to your genome

population.evolve(40, genome => {
  const network = genome.generateNetwork()
  const prediction = network.predict(input)
  // ... return a fitness score according to the accuracy of prediction
})

This will evolve the population, keeping the fittest neural networks according to your fitness function accross 40 generations. The number of iterations is cumulative, this means if your population current generation is 12, evolving with 40 iterations will transform the population to generation 52.

To run this example :

node examples/xor

Sample code :

const { Population } = require('neuroevolution')
const population = new Population(50, 2, 1, false)
const xor = [
  [[0, 0], 0],
  [[0, 1], 1],
  [[1, 0], 1],
  [[1, 1], 0],
]

population.evolve(1000, genome => {
  const network = genome.generateNetwork()
  let error = 0;
  for (const [input, output] of xor) {
    const [prediction] = network.predict(input)
    error += Math.abs(prediction - output)
  }
  return 1 - (error / xor.length)
})

Neuroevolution is MIT licensed.