Arborist: Parallelized, Extensible Random Forests

The Arborist provides a fast, open-source implementation of Leo Brieman's Random Forest algorithm. The Arborist achieves its speed through efficient C++ code and parallel, distributed tree construction.

Bindings are available for R and Python (evolving)

R

The Arborist is available on CRAN as the Rborist package.

Installation of Release Version:

> install.packages('Rborist')

Installation of Development Version:

> ./ArboristBridgeR/Package/Rborist.CRAN.sh
> R CMD INSTALL Rborist_*.*-*.tar.gz

Notes

• Rborist version 0.1-8 now on CRAN.
• Rborist version 0.1-9 under development.

Python

• Some users reporting successful training and prediction.
• Out-of-bag validation NYI.
• Contributors sought.
• Test cases sought.

Performance

Performance metrics will be measured soon using benchm-ml. Partial results can be found here

References

• Scalability Issues in Training Decision Trees (video), Nimbix Developer Summit, 2017.
• Controlling for Monotonicity in Random Forest Regressors (PDF), R in Finance, May 2016.
• Accelerating the Random Forest algorithm for commodity parallel hardware (Video), PyData, July, 2015.
• The Arborist: A High-Performance Random Forest (TM) Implementation, R in Finance, May 2015.
• Training Random Forests on the GPU: Tree Unrolling (PDF), GTC, March, 2015.

News/Changes

• New option 'maxLeaf' to bound the number of terminal (leaf) nodes.
• New option 'autoCompress' to fine-tune sparse internal representation.
• New function 'Streamline' to clear unused fields of trained forest.
• Sparse R-style 'dcgMatrix' format now accepted, with "i/p" encoding.
• Autocompression employed on a per-predictor basis.
• Space-saving 'thinLeaves' option suppresses creation of summary data for narrow workflows.
• Research-inspired option 'splitQuantile' provides fine-tuning of numeric split positions.
• Improved scaling with row count. Correctness errors are being addressed as they are received.