neurotron v2.0.2

Neurotron

A simple library build on brain.js for genetic evolution, NEAT etc.

About

Neurotron uses brain.js NeuralNetwork and provides a couple of useful functions to perform genetic evolution with it.

Installation and Usage

NPM

npm i neurotron

Neurotron is dependent on brain.js, consult https://www.npmjs.com/package/brain.js#Installation-and-Usage for it's information if any problems occur.

Examples

Here's an example, showcasing how to create and use a basic neurotron.

import Neurotron from "neurotron";

/**
 * Creating a neurotron with 5 inputs, 1 hiddenlayer with 5 neurons and 5 outputs
 */
const neurotron = new Neurotron(Neurotron.createNetwork(5, [5], 5))

/**
 * Getting the neurotron's output, given the input [1, 1, 1, 1, 1]
 */
const output = neurotron.getOutput([1, 1, 1, 1, 1])

/**
 * Cloning a neurotron
 */
const neurotronClone = neurotron.clone()

/**
 * Mutating a neurotron with a rate of 0.5
 */
neurotron.mutate(0.5)

Methods

Neurotron.createNetwork(inputSize: number, hiddenLayers: number[], outputSize: number) -> NeuralNetwork

The function to use to create a randomized starting NeuralNetwork, hiddenLayers should specify an array where each number n represents a layer with n neurons.

getOutput(input: number[]) -> output: number[]

The input should have the same length as the specified inputs in the constructor.

clone() -> Neurotron

To clone a neurotron.

cloneNetwork() -> NeuralNetwork

This method gets used internally, and shouldn't be needed.

mutate(rate: number)

Mutates the neurotron directly, where the rate should be between 0 and 1, where 0 means nothing will be mutated and 1 means everything will be mutated fully.

getMutation(base: number, rate: number) -> mutatedBase: number

Mutates a number, based on the rate.

Issues

If you have any issues, you can create a github issue at:
https://github.com/sjoegd/neurotron

Contributors

Fully made by Sjoegd