The non-convex Burer-Monteiro approach works on smooth semidefinite programs

Authors: Nicolas Boumal, Vlad Voroninski, Afonso Bandeira

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

Reproducibility Variable Result LLM Response
Research Type Experimental Interestingly, known algorithms for optimization on manifolds converge to second-order critical points,2 regardless of initialization [Boumal et al., 2016]. ... Indeed, the numerical experiments clearly show that high accuracy solutions can be computed fast using optimization on manifolds, at least for certain applications.
Researcher Affiliation Academia Nicolas Boumal Department of Mathematics Princeton University nboumal@math.princeton.edu Vladislav Voroninski Department of Mathematics Massachusetts Institute of Technology vvlad@math.mit.edu Afonso S. Bandeira Department of Mathematics and Center for Data Science Courant Institute of Mathematical Sciences, New York University bandeira@cims.nyu.edu
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code No The paper mentions "Manopt, a Matlab toolbox for optimization on manifolds. Journal of Machine Learning Research, 15:1455 1459, 2014. URL http://www.manopt.org." as a tool, but it does not state that the specific code for the methodology described in *this* paper is released or available.
Open Datasets No The paper mentions problem types like Max-Cut and community detection, and in Appendix B, a "cycle graph of n=20 nodes" is used for a numerical demonstration. However, it does not provide concrete access information (link, DOI, specific citation with authors/year, or repository) for a public dataset used in experiments.
Dataset Splits No The paper does not specify dataset splits (e.g., training, validation, test percentages or counts) needed to reproduce data partitioning.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory amounts) used for running its experiments.
Software Dependencies No The paper mentions "Manopt [Boumal et al., 2014]", "CVX: Matlab software for disciplined convex programming", and "SDPT3 a MATLAB software package for semidefinite programming", but does not provide specific version numbers for these software dependencies (e.g., "Manopt vX.Y", "CVX 2.1", "SDPT3 4.0").
Experiment Setup Yes We use the Riemannian trust-region method as implemented in Manopt [Boumal et al., 2014], with default parameters for the trust-region subproblem, and stopping criterion gradnorm < 10^-8.