skale-ml

Skale Machine Learning Library (skale-ml) Guide

skale-ml is Skale’s machine learning library. Its goal is to make practical machine learning scalable, easy and accessible to javascript and node.js developers. It consists of common learning algorithms and utilities, including classification, regression and clustering.

skale-ml is currently in alpha mode version.

I - Classification

Logistic regression

Logistic regression is a linear method generally used to predict binary as well ass multiclass responses.

For now, skale-ml only support the binary logistic regression. The generelization into multinomial logistic regression needs to be implemented.

Sample app

In the example folder you will find a sample skale application manipulating the logistic regression model.

First, if not already done, install globally skale toolbelt:

(sudo) npm install -g skale

Then clone skale-ml locally on your laptop using the following command

git clone https://github.com/skale-me/skale-ml.git

Navigate to the logistic regression example application folder, intall dependencies and run the app.

cd /path/to/skale-ml/examples/logreg
npm install
skale run

What's happening ?

This example requires skale-engine and skale-ml.

var skale = require('skale-engine');
var ml = require('skale-ml');

We create a skale context with:

var sc = skale.context();

And generate a random Support Vector Machine dataset, making it persistent in memory to accelerate model training:

var points = ml.randomSVMData(sc, nObservations, nFeatures, seed).persist();

We then instantiate a logistic regression model associated to the previously created dataset:

var model = new ml.LogisticRegressionWithSGD(points);

Finally we train the model on a given number of iterations, display the model weights and end the skale session:

model.train(nIterations, function() {
	console.log(model.weights);
	sc.end();
});

Linear Support Vector Machines (SVMs)

To be documented

---

II - Regression

Linear least squares, Lasso, and ridge regression

To be documented

III - Clustering

K-means

To be documented