Simple-Neural-Net-Library

A simple neural network library written in javascript with multi-layer support.

Introduction

This is a simple neural network library supporting multi-layer fully connected networks.

Network Specifications

Layers

Currently, only fully-connected layers are supported.

Activation Functions

Currently, sigmoid function σ(x) = 1/(1+e^-x) is used. More functions will be added soon.

Error Functions

Currently, Mean Squared Error, M.S.E = 1/2 * (outputs - targets)^2 is used. More will be added soon.

Usage

A new instance of neural network is created as follows:

let nn = new NeuralNetwork(depth,nodeCounts,learningRate);

It takes three parameters:

depth : The number of layers in the neural network. (Number)

nodeCounts : An array of depth number of Number elements representing number of nodes in the layers. (Array)

learningRate : The rate of learning of the neural network. (Number)

To get the output of the neural network, call nn.feedForward(inputs where inputs is the inputs array.

To train the network, call nn.train(expectedOutputs,[inputs]). If inputs is specified, it runs the feedForward and then trains. Otherwise, it uses the existing values (i.e., values from the last time it ran feedForward) to train the network.