Visualizing High-Dimensional Data: MDS, PCA, t-SNE, and UMAP

Presenter: Brian Luke, PhD (ABCS)

Visualizing high-dimensional data presents unique challenges. While Principal Component Analysis (PCA) is a widely used approach, several powerful alternatives exist including Multi-Dimensional Scaling (MDS), t-Distributed Stochastic Neighbor Embedding (t-SNE), and Uniform Manifold Approximation and Projection (UMAP). This demo-based session provided practical examples of generating each of these visualizations using R/RStudio.

Complementary Presentation

For a conceptual understanding of these methods, check out the following ABCS Statistics for Lunch seminar:

• Visualizing High-Dimensional Data: MDS, PCA, t-SNE, and UMAP

R Script

The R Script used for this tutorial can be found here.