/arf_paper

Code and materials to reproduce adversarial RF paper

Primary LanguagePython

Adversarial Random Forests

Code and materials to reproduce paper "Adversarial Random Forests for Density Estimation and Generative Modelling", to appear in Proceedings of the 26th International Conference on Artificial Intelligence and Statistics. Preprint: https://arxiv.org/abs/2205.09435.

The proposed method is implemented in the arf R package, available on CRAN and GitHub. To install the package from CRAN, run:

install.packages("arf")

To install the development version from GitHub using devtools, run:

devtools::install_github("bips-hb/arf")

Directories included in this repository:

  • simulation: Directory containing code to reproduce visual examples and comparison with other tree-based approaches (Sec. 5.1)
  • density_benchmark: Directory containing code to reproduce comparison with alternative PCs in density estimation (Sec. 5.2)
  • generative_benchmark: Directory containing code to reproduce comparison with deep learning approaches in generative modeling (Sec. 5.2-5.3)
  • appx_mnist: Directory containing code to reproduce comparison with conditional GAN for MNIST28 (Appx. B.5)