Notice: This website is continuously updated as new conference proceedings and journal volumes are published. Consequently, the figures presented here may differ from those reported in our publications.
A brief overview of our methodology is provided on the Methods page, along with references to publications documenting our approach and links to the code and data used in each study.
The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates.
Coherence-free Entrywise Estimation of Eigenvectors in Low-rank Signal-plus-noise Matrix Models
Authors: Hao Yan, Keith Levin
NeurIPS 2024 | Venue PDF | LLM Run Details | Input Tokens: 47,585 Total number of tokens sent to the LLM as input for this paper's analysis. | Output Tokens: 4,267 Total number of tokens produced by the LLM (including reasoning/thinking tokens) for this paper's analysis.
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 5 Numerical experiments. We turn to a brief experimental exploration of our theoretical results. |
| Researcher Affiliation | Academia | Hao Yan Department of Statistics University of Wisconsin Madison Madison, WI 53706 United States of America EMAIL. Keith Levin Department of Statistics University of Wisconsin Madison Madison, WI 53706 United States of America EMAIL |
| Pseudocode | Yes | Algorithm 1 Coherence-free eigenvector estimation algorithm. Input: Observed matrix Y Rn n; leading eigenvalue estimate bĪ»; parameter Ī² > 0. Output: bu Rn |
| Open Source Code | Yes | in addition to the algorithmic descriptions in Sections 2 and 3 and the details in Section 5, we have included code for running all reported experiments in our supplemental materials. |
| Open Datasets | No | We consider three distributions for the entries of W : Gaussian, Laplacian and Rademacher, all scaled to have variance Ļ2 = 1. We consider two approaches to generating u . |
| Dataset Splits | No | We report the mean of 20 independent trials for each combination of problem size n, magnitude a and methods for generating u and W . |
| Hardware Specification | No | All experiments were run in a distributed environment on commodity hardware without GPUs. In total, the experiments reported below used 3425 compute-hours. Mean memory usage was 3.5 GB, with a maximum of 11 GB. |
| Software Dependencies | No | The paper mentions 'open source software tools' in the NeurIPS checklist but does not provide specific software or library version numbers. |
| Experiment Setup | Yes | We take Ī» = n log n in all experiments, matching the rate in Remark 1. We consider three distributions for the entries of W : Gaussian, Laplacian and Rademacher, all scaled to have variance Ļ2 = 1. |