SPORF

SPORF (Sparse Projection Oblique Randomer Forests) combines sparse random projections with the random forest algorithm to achieve high accuracy on a variety of datasets.

Currently available in Python and R (not actively developed), SPORF is supported on Linux and Mac OS (and Windows via WSL).

SPORF is optimized for both speed and memory performance through native implementation and multicore parallelization.

Available in a pre-built Gigantum project.

Publications
  1. T. M. Tomita, J. Browne, C. Shen, J. Chung, J. L. Patsolic, B. Falk, J. Yim, C. E. Priebe, R. Burns, M. Maggioni, and J. T. Vogelstein. Sparse Projection Oblique Randomer Forests. arXiv, 2019.
  2. R. Guo, C. Shen, and J. T. Vogelstein. Estimating Information-Theoretic Quantities with Random Forests. arXiv, 2019.
  3. M. Madhyastha, P. Li, J. Browne, V. Strnadova-Neely, C. E. Priebe, R. Burns, and J. T. Vogelstein. Geodesic Learning via Unsupervised Decision Forests. arXiv, 2019.
  4. J. Browne, T. M. Tomita, D. Mhembere, R. Burns, and J. T. Vogelstein. Forest Packing: Fast, Parallel Decision Forests. SIAM International Conference on Data Mining, 2018.
  5. T. Tomita, M. Maggioni, and J. Vogelstein. ROFLMAO: Robust oblique forests with linear MAtrix operations. Proceedings of the 17th SIAM International Conference on Data Mining, SDM 2017, 2017.