https://github.com/sircamp/simple-extreme-learning-machine

A Simple Extreme Learning Machine implementation in Scala With Breeze
https://github.com/sircamp/simple-extreme-learning-machine

datascience deep-learning extreme-learning-machine machine-learning scala

Last synced: 9 days ago
JSON representation

A Simple Extreme Learning Machine implementation in Scala With Breeze

• Host: GitHub
• URL: https://github.com/sircamp/simple-extreme-learning-machine
• Owner: sirCamp
• License: mit
• Created: 2019-08-19T19:59:17.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2019-08-21T06:40:57.000Z (over 5 years ago)
• Last Synced: 2024-11-19T13:46:04.551Z (2 months ago)
• Topics: datascience, deep-learning, extreme-learning-machine, machine-learning, scala
• Language: Scala
• Size: 125 KB
• Stars: 3
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
[![Build Status](https://travis-ci.com/sirCamp/simple-extreme-learning-machine.svg?branch=master)](https://travis-ci.com/sirCamp/simple-extreme-learning-machine)

[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)

[![Scala](https://img.shields.io/badge/scala-v2.12-blue)](https://img.shields.io/badge/scala-v2.12-blue)

# A simple implementation of Extreme Learning Machine

This is a simple scala & Breeze implementation of an Extreme Learning Machine.

This library is designed to have great performances in scientific computation thanks to the linear algebra optimizations.

### How to

In order to use take a look to the following example:

```scala

 import com.sircamp.elm.ExtremeLearningMachine

 var featuresLength = 28*28 //MINST dataset

 var hiddenLayerDimension = 1024

 val elm = new ExtremeLearningMachine(featuresLength, hiddenLayerDimension)

 

 /**

  Initialize the weights of the hidden layer with random uniform distribution

  Otherwise you can set the weights by your own. 

  Weights must be a DenseMatrix[Double] where rows are equal to the featuresLength

 **/

 elm.initializeWeights() 

 /**

  fit the model.

  XTrain and yTrain must be DenseMatrix[Double].

  yTrain must be the one hot encoded version of the original label

**/

 elm.fit(XTrain, yTrain)

/**

  predictClasses return a DenseVector[Int] with the index of the predicted class 

**/

 var yPred = elm.predictClasses(XTest)

/**

  predict return a DenseMatrix[Double] where each row contains the probability of the element to belongs to the class 

**/

 var yProbabilityPred = elm.predict(XTest)

 println("Accuracy: "+Metrics.accuracy_score(yPlain,yPred))

```

Set a different Activation function

```scala

import com.sircamp.elm.ExtremeLearningMachine

 var featuresLength = 28*28 //MINST dataset

 var hiddenLayerDimension = 1024

 val elm = new ExtremeLearningMachine(featuresLength, hiddenLayerDimension)

 

 /**

  Initialize the weights of the hidden layer with random uniform distribution

  Otherwise you can set the weights by your own. 

  Weights must be a DenseMatrix[Double] where rows are equal to the featuresLength

 **/

 elm.initializeWeights() 

 /**

   This return the LeakyReLu function with the alpha param

 **/

 elm.setActivationFunction(ActivationFunctions.leakyReLu(0.2))

 /**

   This return the Tanh function

 **/

 elm.setActivationFunction(ActivationFunctions.tanh)

```

For more example take a look to the tests