songR: Self-Organizing Nebulous Growths for Dimensionality Reduction

A SONG-inspired tool for nonlinear dimensionality reduction and data visualization in R, implementing the Self-Organizing Nebulous Growths (SONG) algorithm of Senanayake et al. (2021) doi:10.1109/TNNLS.2020.3023941 natively in R and C++ and staying as close to the reference implementation as is feasible across the two ecosystems. SONG is a parametric method that supports incremental addition of new data to existing visualizations without reinitialization, handles both homogeneous and heterogeneous data increments, and is robust to noise and highly mixed clusters. Clustering quality matches the reference closely (validated by reference-parity tests); the default visualization is close in global structure but not identical in absolute layout. Includes an interactive Shiny application for comparing SONG, t-SNE, and UMAP visualizations side-by-side.

References

Senanayake, D. A., Wang, W., Naik, S. H., & Halgamuge, S. (2021). Self-Organizing Nebulous Growths for Robust and Incremental Data Visualization. IEEE Transactions on Neural Networks and Learning Systems, 32(10), 4588–4602. doi:10.1109/TNNLS.2020.3023941

See also

Useful links:

• https://cttir.github.io/songR/

• https://github.com/cttir/songR

• Report bugs at https://github.com/cttir/songR/issues

Author

Maintainer: R. Heller raban.heller@uni-ulm.de (ORCID) [copyright holder]

Authors:

• R. Heller raban.heller@uni-ulm.de (ORCID) [copyright holder]

Other contributors:

• Damith A. Senanayake (Author of the original SONG algorithm and its BSD-3-Clause Python implementation; see inst/COPYRIGHTS) [copyright holder]