Enhancing Sufficient Dimension Reduction via Hellinger Correlation

Authors: Seungbeom Hong, Ilmun Kim, Jun Song

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Through extensive numerical experiments, we demonstrate that our proposed method significantly enhances and outperforms existing SDR methods. In Section 4, we provide simulation results that compare our method with existing ones. Section 5 presents a real data application of the method.
Researcher Affiliation Academia 1Department of Statistics, Korea University, Seoul, South Korea 2Department of Applied Statistics, Yonsei University, Seoul, South Korea.
Pseudocode No The paper describes the method in prose but does not include any explicitly labeled pseudocode or algorithm blocks.
Open Source Code Yes The code that implements our proposed method is available at https://github.com/JSong Lab/SDR HC.
Open Datasets Yes We apply our methods to the real estate valuation dataset in the UCI Machine Learning Repository (Yeh, 2018).
Dataset Splits No Subsequently, we randomly divide the dataset into a training sample of size 300 and use the remaining objects as the test sample. (No explicit mention of a validation split.)
Hardware Specification No The paper does not specify any hardware details (e.g., CPU, GPU models, memory) used for running the experiments.
Software Dependencies No The paper does not list specific software dependencies with version numbers (e.g., Python 3.x, PyTorch x.x).
Experiment Setup Yes To assess the performance of our method, we employ the following metric to quantify the difference between two subspaces: (ST rue, SEstimated) = PST rue PSEstimated , where is the maximum eigenvalue of a matrix and PST rue and PSEstimated are the orthogonal projection matrices of the subspace ST rue = Span(η ) and SEstimated = Span(ˆη). A smaller value of indicates a more accurate estimation. Subsequently, we randomly divide the dataset into a training sample of size 300 and use the remaining objects as the test sample. SDR methods are then applied to the training set to extract the SDR direction, followed by fitting a local polynomial regression using the remaining variables to predict the house price. The weights are given equally for each observation and quadratic polynomial was used to fit the model.