Visualizing High-Dimensional Data: MDS, PCA, t-SNE, and UMAP
To Know
Where:
Bldg 549, Frederick, Ft. Detrick, Executive Board Room
Organizer:
ABCS/FNLCR
Presented By:
Brian Luke (Advanced Biomedical Computational Science, ABCS)
About this Class
Visualizing high-dimensional data can be problematic. A common method is Principal Component Analysis (PCA), but other methods include Multi-Dimensional Scaling (MDS), t-Distributed Stochastic Neighbor Embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP). This talk discusses the differences between these methods without going into complicated mathematics and will hopefully allow you to determine which type of plot is best for your data.