Optimal Non-Convex Exact Recovery in Stochastic Block Model via Projected Power Method

Authors: Peng Wang, Huikang Liu, Zirui Zhou, Anthony Man-Cho So

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We also present numerical results of the proposed method to support and complement our theoretical development.
Researcher Affiliation Collaboration 1Department of Systems Engineering and Engineering Management, The Chinese University of Hong Kong, Hong Kong 2Business School, Imperial College London, London, United Kingdom 3Huawei Technologies Canada Co., Ltd., Burnaby, Canada.
Pseudocode Yes Algorithm 1 Projected Power Method for Solving Problem (MLE)
Open Source Code Yes Our code is available at https://github.com/peng8wang/ICML2021-PPM-SBM.
Open Datasets Yes We use the data sets polbooks, polblogs, and football downloaded from the Suite Sparse Matrix Collection (Davis & Hu, 2011).
Dataset Splits No The paper mentions evaluating performance on synthetic and real datasets but does not explicitly provide information on training, validation, or test splits of these datasets, nor does it refer to standard splits for these specific datasets.
Hardware Specification Yes All of our simulations are implemented in MATLAB R2020a on a PC running Windows 10 with 16GB memory and Intel(R) Core(TM) i5-8600 3.10GHz CPU.
Software Dependencies Yes All of our simulations are implemented in MATLAB R2020a
Experiment Setup Yes The stopping criteria for the tested methods are set as follows. For PPM, we terminate it when there exists some iterate k 6 such that Hk Hl F 10 3 for some k 5 l k 1; for ADMM, we terminate it when the norm of difference of two consecutive iterates is less than 10 3. ... Moreover, we set the maximum iteration number for PPM and ADMM as 1000.