🧬 NEAT.js

A TypeScript implementation of the NEAT (NeuroEvolution of Augmenting Topologies) algorithm for evolving neural networks.

🎮 Live Demo

Try the interactive demo: https://leoni4.github.io/neat-js/

The demo showcases NEAT solving various problems including XOR, parity functions, and classification tasks with real-time visualization of the neural network evolution.

Installation

npm install @leoni4/neat-js

Usage

Basic Library Usage

import { Neat } from '@leoni4/neat-js';

// Create a NEAT instance
const neat = new Neat(
    2, // input nodes
    1, // output nodes
    1000, // population size
);

// Training data (XOR example)
const inputs = [
    [0, 0],
    [1, 1],
    [1, 0],
    [0, 1],
];
const outputs = [[0], [0], [1], [1]];

const solved = neat.fit(inputs, outputs, {
    verbose: 1,
});

console.log('solved in:', solved.epochs);

Manual Library Usage

Could be helpfull with non linear score determination, or complex environment

import { Neat } from '@leoni4/neat-js';

// Create a NEAT instance same as above
const neat = new Neat(2, 1, 1000);

// Training data (XOR example) ... same as above
const inputs = [
    [0, 0],
    [1, 1],
    [1, 0],
    [0, 1],
];
const outputs = [[0], [0], [1], [1]];

let finalError = 1; // just for the demo
let counter = 0;
// Evolution loop
for (let epoch = 0; epoch < 1000; epoch++) {
    // Evaluate each client
    for (const client of neat.clients) {
        let error = 0;
        inputs.forEach((input, i) => {
            const output = client.calculate(input)[0];
            error += Math.abs(output - outputs[i][0]);
        });
        error = error / inputs.length;

        // calculate error and score of the client
        client.error = error;
        client.score = 1 - client.error;

        if (finalError > error) {
            finalError = error;
        }
    }
    counter++;
    // Evolve to next generation
    neat.evolve();

    if (epoch % 10 === 0) console.log('Epoch:', epoch, ' Error:', finalError);

    if (finalError < 0.01) break;
}

console.log('Total epochs:', counter, ' Final Error:', finalError);
// ... rest of the logic

Development

Running the Demo

Clone the repository and run the demo locally:

git clone https://github.com/leoni4/neat-js.git
cd neat-js
npm install
npm start

The demo will start a Vite development server at http://localhost:3000 with a visual demonstration of the NEAT algorithm solving the selected problem.

Available Scripts

• npm start - Start the Vite development server with demo
• npm run build - Build the library for production
• npm run build:demo - Build the demo for production
• npm test - Run tests
• npm run test:watch - Run tests in watch mode
• npm run lint - Lint the codebase
• npm run typecheck - Run TypeScript type checking

API

Core Classes

• Neat - Main NEAT algorithm class
• Client - Individual neural network in the population
• Genome - Genetic representation of a neural network
• Species - Group of similar genomes

Data Structures

• RandomHashSet - Efficient random selection from a set
• RandomSelector - Weighted random selection

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes with tests
4. Ensure all tests pass: npm test
5. Submit a pull request

License

MIT

Repository

https://github.com/leoni4/neat-js