Scale-invariant Optimal Sampling for Rare-events Data and Sparse Models

Authors: Jing Wang, HaiYing Wang, Hao Zhang

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

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct numerical experiments using both simulated and real-world data sets to demonstrate the performance of the proposed methods.
Researcher Affiliation Academia Jing Wang Department of Statistics University of Connecticut Storrs, CT 06269 jing.7.wang@uconn.edu Hai Ying Wang Department of Statistics University of Connecticut Storrs, CT 06269 haiying.wang@uconn.edu Hao Helen Zhang Department of Mathematics University of Arizona hzhang@math.arizona.edu
Pseudocode Yes Algorithm 1 Poisson Subsampling algorithm
Open Source Code Yes Codes are submitted as supplement for anonymity. They will be released in a public github repository after the review period.
Open Datasets Yes (i) Covtype data set: It is available at https://archive.ics.uci.edu/ml/datasets/ covertype, with N = 581012 observations and 54 covariates...
Dataset Splits Yes We use the 5-fold cross-validation and Bayesian information criterion (BIC) to determine the tuning parameter λ for the lasso and the adaptive lasso, and choose γ = 1 for the adaptive lasso.
Hardware Specification No The paper states that codes were 'implemented on a Linux workstation' in Section 5.1.3, but does not provide specific details on the CPU, GPU models, or memory of the hardware used.
Software Dependencies Yes Our codes are written in the julia programming language [2] and implemented on a Linux workstation. The lasso pathes are solved with Lasso.jl [13].
Experiment Setup Yes We use the 5-fold cross-validation and Bayesian information criterion (BIC) to determine the tuning parameter λ for the lasso and the adaptive lasso, and choose γ = 1 for the adaptive lasso.