On Sparse Canonical Correlation Analysis

Authors: Yongchun Li, Santanu Dey, Weijun Xie

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

Reproducibility Variable Result LLM Response
Research Type Experimental Section 5 numerically test the proposed formulations and algorithms.
Researcher Affiliation Academia Yongchun Li University of Tennessee ycli@utk.edu Santanu S. Dey Georgia Tech santanu.dey@isye.gatech.edu Weijun Xie Georgia Tech wxie@gatech.edu
Pseudocode Yes Algorithm 1 An exact algorithm for SCCA (1) when s1 r and s2 ˆr
Open Source Code Yes The codes and data used in our experiments are available at https://github.com/yongchunli-13/SCCA.git.
Open Datasets Yes The codes and data used in our experiments are available at https://github.com/yongchunli-13/SCCA.git. Also, the paper cites UCI datasets [5] and breast cancer dataset [8], which are commonly publicly available.
Dataset Splits No The paper states "The dataset is split into the first n variables and the remaining m variables to construct the sample covariance matrices A, B, C." in Section 5.1. This refers to a partitioning of variables, not a train/validation/test split for data samples typically used in model evaluation.
Hardware Specification Yes All the experiments are conducted in Python 3.6 with calls to Gurobi 9.5.2 and MOSEK 10.0.29 on a PC with 10-core CPU, 16-core GPU, and 16GB of memory.
Software Dependencies Yes All the experiments are conducted in Python 3.6 with calls to Gurobi 9.5.2 and MOSEK 10.0.29 on a PC with 10-core CPU, 16-core GPU, and 16GB of memory.
Experiment Setup Yes Section 5.1 "Experimental setup" details the generation of synthetic data, including parameters (n, m, s1, s2) and sampling N=5,000 data samples. It also states the time limit for experiments: "the time limit is one hour".