LOL is a robust and powerful for learning lower-dimensional representation in high-dimensional, low sample size setting.
LOL is commercial open-source software and released under Apache-2.0 License.
- J. T. Vogelstein, M. Tang, D. Zheng, R. Burns, and M. Maggioni. Linear Optimal Low Rank Projection Provably Outperforms Principal Components Analysis in High-Dimensional Multi-Class Data. arXiv, 2018.