Partitioning Structure Learning for Segmented Linear Regression Trees

Authors: Xiangyu Zheng, Song Xi Chen

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The practical performance of the SLRT and its ensemble versions are evaluated via numerical simulations and empirical studies. The latter shows their advantageous predictive performance over a set of state-of-the-art tree-based models on well-studied public datasets.
Researcher Affiliation Academia Xiangyu Zheng Peking University zhengxiangyu@pku.edu.cn Song Xi Chen Peking University csx@gsm.pku.edu.cn
Pseudocode Yes Algorithm 1 Recursive Partitioning for Conditional Uncorrelated Regressors; Algorithm 2 Split Selection for Correlated Regressors
Open Source Code Yes The source code of the algorithm is available in the supplementary material.
Open Datasets Yes The predictive performance is tested on 9 benchmark datasets from the Stat Lib library [23] and the UCI Machine Learning Repository [24]
Dataset Splits Yes Then, the optimally pruned subtree is T(α ). Let b L be the number of terminal nodes in T(α ), under certain general conditions for the distribution of ε and given appropriate α , it can be proved that b L converges to the genuine number of segments L0 in probability. ... The optimal complexity parameter α is selected from { αk}K k=1 by the ten-fold cross-validation to optimize the average predictive accuracy measured by the sum of squared residuals.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper mentions general software or methods (e.g., 'LASSO procedure', 'random forests', 'CART') but does not specify any software dependencies with version numbers.
Experiment Setup Yes With the same stopping parameter of Nmin = 10, Depmax = 10, we applied SLRT and CART respectively, obtaining the approximated surface in Figure 4 and 5.