RODRIGUES. M. A.; RODRIGUES, Matheus Andrade.
Resumo:
We applied PCA, t-SNE, and UMAP to datasets from genetic interpretation calculators containing data of Jewish
ethnic groups, various non-Jewish neighbors, and correlated ethnicities, using the R software. We conducted a
visual comparison of the generated results and used microbenchmarking to measure the execution time of the
methods. t-SNE and UMAP are efficient for working with local aspects of visualization, while PCA is suitable when
the number of samples is small. t-SNE and UMAP are capable of forming clusters that would not be seen using
PCA alone. However, they are slower than PCA, and the visualizations generated by them change when the
algorithm is run again.