JellyBrain is a simple neural network written in Javascript. This was written as an exercise to learn how neural networks work.
const JellyBrain = require('../JellyBrain.js');
let brain = new JellyBrain(2, 2, 1); // 2 inputs, 2 hidden nodes, 1 output
brain.train([0.2, 0.5], 1);
brain.guess([0.1, 0.6]);It is possible to use a variety of cost functions and activation functions. It is also possible to train in batches, although the implementation avoids using tensors to simplify things.
const {JellyBrain, costFuncs, activationFuncs} = require('../JellyBrain.js');
// example parameters for training on the mnist dataset
let brain = new JellyBrain(784, 784, 10, costFuncs.crossEntropy, 0.001, activationFuncs.sigmoid, activationFuncs.softmax);
let simpleBrain = new JellyBrain(2, 2, 1);
simpleBrain.addToBatch([0.2, 0.5], 1);
simpleBrain.addToBatch([0.6, 0.4], 0.7);
simpleBrain.addToBatch([0.1, 0.2], 0.2);
simpleBrain.computeBatch();
simpleBrain.guess([0.1, 0.6]);