Polynomial time guarantees for the Burer-Monteiro method

Authors: Diego Cifuentes, Ankur Moitra

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 6 Experiments We present some experimental results to complement our theorems. We rely on the library NLopt [24] (MIT license) to solve the nonlinear problem (BM). Concretely, we use the augmented Lagrangian method (ALM) implemented in NLopt (which is based on [8]), and we use the preconditioned truncated Newton method as the subroutine. We also rely on the commercial solver Mosek for SDPs. Figure 1 shows the percentage of experiments solved correctly for each value of r and p.
Researcher Affiliation Academia Diego Cifuentes School of Industrial and Systems Engineering Georgia Institute of Technology Atlanta, GA 30332 diego.cifuentes@isye.gatech.edu Ankur Moitra Department of Mathematics Massachusetts Institute of Technology Cambridge, MA 02139 moitra@mit.edu
Pseudocode No The paper refers to algorithms (e.g., 'Theorem 4 below we show that AFAC points can be computed in polynomial time'), but it does not provide explicit pseudocode or an algorithm block for any method.
Open Source Code Yes Did you include the code, data, and instructions needed to reproduce the main experimental results (either in the supplemental material or as a URL)? [Yes] In supplementary material
Open Datasets No For our first experiment we consider a random SDP with a planted solution. More precisely, we take a matrix X0 S50, X0 0 of rank r {4, 7, 12}, and generate a random SDP for which X0 is an optimal solution. To do so, we generate m := τ(r) random constraints that are satisfied at X0, and then find a cost matrix C in the normal cone of X0. We generate 100 random SDPs for each r. We use random data for the experiments.
Dataset Splits No The paper states 'N/A' for specifying 'training details (e.g., data splits, hyperparameters, how they were chosen)' in the ethics review guidelines.
Hardware Specification No The paper states 'N/A' for 'total amount of compute and the type of resources used (e.g., type of GPUs, internal cluster, or cloud provider)' in the ethics review guidelines, indicating no specific hardware details.
Software Dependencies No The paper mentions 'NLopt [24]' and 'Mosek' as software used, but it does not specify version numbers for these components.
Experiment Setup No The paper states 'N/A' for specifying 'training details (e.g., data splits, hyperparameters, how they were chosen)' in the ethics review guidelines.