Centroids

Explainability framework for dimensionality reduction plots in neurogenetic datasets.

Centroids improves interpretability for dimensionality-reduction outputs used in biomedical machine learning.

Project scope:

  • Implemented PCA, t-SNE, and UMAP pipelines for high-dimensional neurogenetic data analysis.
  • Added feature-importance-driven ranking and centroid overlays for more interpretable embedding views.
  • Automated large parameter sweeps for reduction methods to accelerate reproducible experimentation.

This work contributed to our IEEE CIBCB 2024 publication, which received the Best Paper Award.