PaCMAP
Dimensionality reduction that preserves both local neighborhoods and global structure.
PaCMAP (Pairwise Controlled Manifold Approximation Projection) maps high-dimensional data into a low-dimensional space for visualization. It balances nearby, mid-near, and further pairs to preserve local neighborhoods and the broader shape of the data.
Developed with Yingfan Wang, Cynthia Rudin, and Yaron Shaposhnik, this work appeared in the Journal of Machine Learning Research in 2021.