Dimensionality Reduction

keywords: ALFA, dimensionality, manifold

A lot of big datasets reside in a very high dimensional space. These datasets can be visualized in a two or three dimensional space using dimensionality reduction algorithms. Dimensionality reduction algorithms transform high dimensional datasets into a lower dimensional space for visualization and analysis. Dimensionality reduction algorithms can be linear or nonlinear depending on the intrinsic structure of the data.

ALFA currently supports the following dimensionality reduction algorithms:

Linear

1. Principal Component Analysis

Nonlinear

1. Isomap
2. Locally Linear Embedding
3. t-SNE

The following example demonstrates how to visualize high dimensional datasets using dimensionality reduction algorithms in ALFA.

run Isomap on [dataset]

• The output is displayed as a two dimensional scatterplot.
• The parameters of the algorithm can be modified using the editable panel on the right.