/dimred

Dimensionality reduction methods

Primary LanguageJavaScript

Dimensionality reduction tools for browsers and Node.js

The dimred package is a wrapper around some dimensionality reduction methods implemented in JavaScript. It simplifies their API and makes it possible to try different algorithms without adapting data and code for each case. You can also use dimred via CLI to generate a lower-dimensional representation of a dataset without writing codeat all.

Supported dimensionality reduction methods

  • PCA Princinpal Component Analysis (pca)
  • SOM Self-Organizing Map (som)
  • tSNE (tsne)
  • UMAP (umap)
  • Autoencoder (ae or autoencoder)

Example

const mkdata = require('mkdata')
const dimred = require('dimred')

// Generate a dataset with 1000 samples
// X: Array (1000, 10)
const [X, _] = mkdata.friedman1({
  'nSamples': 1000
}) 

// Run dimensionality reduction
// emb: Array (1000, 2)
const emb = dimred(X, {
  'method': 'pca',
  'dims': 2
})

Web demo

All methods included in the dimred package are available online on StatSim Vis