Exact Recovery of Hard Thresholding Pursuit

Authors: Xiaotong Yuan, Ping Li, Tong Zhang

NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Numerical results on simulated data confirms our theoretical predictions. In this section, we conduct a group of Monte-Carlo simulation experiments on sparse linear regression and sparse logistic regression models to verify our theoretical predictions.
Researcher Affiliation Academia Xiao-Tong Yuan B-DAT Lab Nanjing University of Info. Sci.&Tech. Nanjing, Jiangsu, 210044, China xtyuan@nuist.edu.cn Ping Li Tong Zhang Depart. of Statistics and Depart. of CS Rutgers University Piscataway, NJ, 08854, USA {pingli,tzhang}@stat.rutgers.edu
Pseudocode Yes Algorithm 1: Hard Thresholding Pursuit.
Open Source Code No The paper does not provide any explicit statements or links indicating that its source code is publicly available.
Open Datasets No We consider a synthetic data model in which the sparse parameter w is a p = 500 dimensional vector that has k = 50 nonzero entries drawn independently from the standard Gaussian distribution.
Dataset Splits No The paper describes generating synthetic data and varying sample size 'n', but does not explicitly state train/validation/test dataset splits or their proportions.
Hardware Specification No The paper does not specify the hardware (e.g., CPU, GPU models, memory) used for running the experiments.
Software Dependencies No The paper does not specify any software dependencies (e.g., library names with version numbers) used for the experiments.
Experiment Setup Yes Data generation: We consider a synthetic data model in which the sparse parameter w is a p = 500 dimensional vector that has k = 50 nonzero entries drawn independently from the standard Gaussian distribution. Each data sample u is a normally distributed dense vector. For the linear regression model, the responses are generated by v = u w + ε where ε is a standard Gaussion noise. For the logistic regression model, the data labels, v { 1, 1}, are then generated randomly according to the Bernoulli distribution P(v = 1|u; w) = exp(2 w u)/(1 + exp(2 w u)). We allow the sample size n to be varying and for each n, we generate 100 random copies of data independently. In our experiment, we test HTP with varying sparsity level k k.