Bregman Alternating Direction Method of Multipliers

Authors: Huahua Wang, Arindam Banerjee

NeurIPS 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we use BADMM to solve the mass transportation problem [18]:
Researcher Affiliation Academia Huahua Wang, Arindam Banerjee Dept of Computer Science & Engg, University of Minnesota, Twin Cities {huwang,banerjee}@cs.umn.edu
Pseudocode No The paper provides algorithmic steps through mathematical equations (e.g., (7)-(9), (11)-(13)) but does not include a clearly labeled 'Pseudocode' or 'Algorithm' block.
Open Source Code Yes GPU code is available on https://github.com/anteagle/GPU_BADMM_MT
Open Datasets No The paper uses the mass transportation problem where data (cost matrix C) is 'randomly generated from the uniform distribution', not a pre-existing, publicly available dataset with concrete access information.
Dataset Splits No The paper describes generating data (cost matrix C is randomly generated) but does not provide specific training, validation, or test dataset split percentages or counts.
Hardware Specification Yes For comparison, BADMM is run in parallel on a Tesla M2070 GPU with 5G memory and 448 cores1. We run Gurobi on two settings: a Mac laptop with 8G memory and a server with 86G memory, respectively.
Software Dependencies No The paper mentions 'Gurobi' as commercial software and 'LP solvers' but does not provide specific version numbers for these or any other software dependencies.
Experiment Setup Yes We choose the best parameter for BADMM (ρ = 0.001) and ADMM (ρ = 0.001). The stopping condition is either when the number of iterations exceeds 2000 or when the primal-dual residual is less than 10 4.