Stability of Random Forests and Coverage of Random-Forest Prediction Intervals

Authors: Yan Wang, Huaiqing Wu, Dan Nettleton

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical results show that stability may persist even beyond our assumption and hold for heavy-tailed Y 2. and Numerically show that RF stability may hold beyond the above light-tail assumption;
Researcher Affiliation Academia Yan Wang Department of Mathematics Wayne State University Detroit, MI 48202 wangyan@wayne.edu Huaiqing Wu, Dan Nettleton Department of Statistics Iowa State University Ames, IA 50011 {isuhwu,dnett}@iastate.edu
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide concrete access to source code for the methodology described in this paper. It mentions using the "randomForest" package in R, which is a third-party tool, but does not provide its own implementation code.
Open Datasets No We created a virtual dataset consisting of n 4000 points. We let Y be a standard Cauchy random variable, which is even without a well-defined mean. The feature vector X P R3 is determined as X r0.5Y sinp Y q, Y 2 0.2Y 3, It Y ą 0u ζs T where ζ is a standard normal random variable. The paper describes how the data was generated but does not provide access information (link, DOI, repository) for this virtual dataset.
Dataset Splits Yes We used 3000 of the points for training and 1000 of them as test points.
Hardware Specification No The paper only states: 'The computation can be done within a few minutes on a laptop.' This is not specific enough to identify the hardware.
Software Dependencies No The paper mentions using the "random Forest package in R" but does not provide specific version numbers for the software dependencies.
Experiment Setup Yes Using the random Forest package with default setting (except letting B 1000), we had an output RF predictor RFB.