Accelerated Primal-Dual Methods for Convex-Strongly-Concave Saddle Point Problems

Authors: Mohammad Khalafi, Digvijay Boob

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We verify our findings with numerical experiments.
Researcher Affiliation Academia 1Department of Operations Research and Engineering Management, Southern Methodist University, Dallas TX, USA.
Pseudocode Yes Algorithm 1 Linearized PD (LPD) method; Algorithm 2 Accelerated Linearized PD (ALPD) method; Algorithm 3 Inexact ALPD Method
Open Source Code No The paper does not provide a specific statement or link indicating the release of open-source code for the described methodology.
Open Datasets No We set f(x) = 1/2x Qx + c x as a convex quadratic function where Q Rn n is a randomly generated positive semidefinite matrix and c Rn is a random vector. We also generate matrix A Rm n and b Rm randomly. For these experiments, we set the penalty parameter ρ = 1 and m = n = 100. Appendix D provides detailed information on the exact functions used for the random number generation. We set Lf = 200 since eigenvalues of Q are generated uniformly on [0, 200]. The paper describes how synthetic data is generated but does not provide access to the specific datasets used in the experiments, nor does it cite publicly available benchmark datasets used in their experimental setup.
Dataset Splits No The paper describes generating synthetic data for numerical experiments but does not explicitly mention training, validation, or test dataset splits.
Hardware Specification Yes All experiments are performed on 64-bit Windows 10 with Intel i5-9500U @3.00GHz and 16GB RAM.
Software Dependencies No The paper does not specify any software dependencies with version numbers (e.g., programming languages, libraries, frameworks).
Experiment Setup Yes We set f(x) = 1/2x Qx + c x as a convex quadratic function where Q Rn n is a randomly generated positive semidefinite matrix and c Rn is a random vector. We also generate matrix A Rm n and b Rm randomly. For these experiments, we set the penalty parameter ρ = 1 and m = n = 100. Appendix D provides detailed information on the exact functions used for the random number generation. We set Lf = 200 since eigenvalues of Q are generated uniformly on [0, 200].