Adversarial random forests for density estimation and generative modeling
This provides a faster alternative for density estimation and generative modeling in tabular data applications, though it appears incremental as it adapts existing concepts to tree-based methods.
The authors tackled density estimation and data synthesis for tabular data by proposing adversarial random forests, achieving comparable or superior performance to state-of-the-art models while executing about two orders of magnitude faster on average.
We propose methods for density estimation and data synthesis using a novel form of unsupervised random forests. Inspired by generative adversarial networks, we implement a recursive procedure in which trees gradually learn structural properties of the data through alternating rounds of generation and discrimination. The method is provably consistent under minimal assumptions. Unlike classic tree-based alternatives, our approach provides smooth (un)conditional densities and allows for fully synthetic data generation. We achieve comparable or superior performance to state-of-the-art probabilistic circuits and deep learning models on various tabular data benchmarks while executing about two orders of magnitude faster on average. An accompanying $\texttt{R}$ package, $\texttt{arf}$, is available on $\texttt{CRAN}$.