gomind

README

GoMind

Installation

go get github.com/surenderthakran/gomind

About

GoMind is a neural network library written entirely in Go. It only supports a single hidden layer (for now). The network learns from a training set using back-propagation algorithm.

Some of the salient features of GoMind are:

• Supports following activation functions:
  • Linear (Default)
  • Sigmoid
  • ReLU
  • Leaky ReLU
• Smartly estimates ideal number of hidden layer neurons (if a count is not given during model configuration) for given input and output sizes.
• Uses Mean Squared Error function to calculate error while back propagating.

Note: To understand the basic functioning of back-propagation in neural networks, one can refer to my blog here.

Usage

package main

import (
	"github.com/singhsurender/gomind"
)

func main() {
	trainingSet := [][][]float64{
		[][]float64{[]float64{0, 0}, []float64{0}},
		[][]float64{[]float64{0, 1}, []float64{1}},
		[][]float64{[]float64{1, 0}, []float64{1}},
		[][]float64{[]float64{1, 1}, []float64{0}},
	}

	mind, err := gomind.New(&gomind.ModelConfiguration{
		NumberOfInputs:                    2,
		NumberOfOutputs:                   1,
		NumberOfHiddenLayerNeurons:        16,
		HiddenLayerActivationFunctionName: "relu",
		OutputLayerActivationFunctionName: "sigmoid",
	})
	if err != nil {
		return nil, fmt.Errorf("unable to create neural network. %v", err)
	}

	for i := 0; i < 500; i++ {
		rand := rand.Intn(4)
		input := trainingSet[rand][0]
		output := trainingSet[rand][1]

		if err := mind.LearnSample(input, output); err != nil {
			mind.Describe(true)
			return nil, fmt.Errorf("error while learning from sample input: %v, target: %v. %v", input, output, err)
		}

		actual := mind.LastOutput()
		outputError, err := mind.CalculateError(output)
		if err != nil {
			mind.Describe(true)
			return nil, fmt.Errorf("error while calculating error for input: %v, target: %v and actual: %v. %v", input, output, actual, err)
		}

		outputAccuracy, err := mind.CalculateAccuracy(output)
		if err != nil {
			mind.Describe(true)
			return nil, fmt.Errorf("error while calculating error for input: %v, target: %v and actual: %v. %v", input, output, actual, err)
		}

		fmt.Println("Index: %v, Input: %v, Target: %v, Actual: %v, Error: %v, Accuracy: %v\n", i, input, output, actual, outputError, outputAccuracy)
	}
}

API Documentation

The documentation for various methods exposed by the library can be found at: https://godoc.org/github.com/surenderthakran/gomind

Documentation

Overview

Package gomind for a simple Multi Layer Perceptron (MLP) Feed Forward Artificial Neural Network library.

Index

• type Model
  • func New(configuration *ModelConfiguration) (*Model, error)
  • func (model *Model) CalculateError(targetOutput []float64) (float64, error)
  • func (model *Model) Describe(showNeurons bool)
  • func (model *Model) LastOutput() []float64
  • func (model *Model) LearnSample(input, output []float64) error
• type ModelConfiguration
• type NeuralNetworkInterface

Types

type Model

type Model struct {
	// contains filtered or unexported fields
}

Model type defines the neural network's architecture and metadata.

func New

func New(configuration *ModelConfiguration) (*Model, error)

New is used to create a new GoMind network model.

func (*Model) CalculateError

func (model *Model) CalculateError(targetOutput []float64) (float64, error)

CalculateError function generates the error value for the given target output against the network's last output.

func (*Model) Describe

func (model *Model) Describe(showNeurons bool)

Describe function prints the current state of the neural network and its components.

func (*Model) LastOutput

func (model *Model) LastOutput() []float64

LastOutput function returns the last output of the network.

func (*Model) LearnSample

func (model *Model) LearnSample(input, output []float64) error

LearnSample function trains the neural network using the given input/output sample.

type ModelConfiguration

type ModelConfiguration struct {
	NumberOfInputs                    int // mandatory
	NumberOfOutputs                   int // mandatory
	NumberOfHiddenLayerNeurons        int
	LearningRate                      float64
	HiddenLayerActivationFunctionName string
	OutputLayerActivationFunctionName string
}

ModelConfiguration type defines the network configuration template filled by external code while creating a new model.

type NeuralNetworkInterface

type NeuralNetworkInterface interface {
	CalculateOutput(input []float64) []float64
	LastOutput() []float64
	HiddenLayer() *layer.Layer
	OutputLayer() *layer.Layer
	CalculateNewOutputLayerWeights(outputs, targetOutputs []float64) error
	CalculateNewHiddenLayerWeights() error
	UpdateWeights()
	CalculateError(targetOutput []float64) (float64, error)
}

NeuralNetworkInterface defines methods used by gomind from network.NeuralNetwork's type.