Accelerated Quasi-Newton Proximal Extragradient: Faster Rate for Smooth Convex Optimization

Authors: Ruichen Jiang, Aryan Mokhtari

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

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we compare the numerical performance of our proposed A-QNPE method with NAG and the classical BFGS quasi-Newton method. Figure 1: Numerical results for logistic regression on a synthetic dataset.
Researcher Affiliation Academia Ruichen Jiang ECE Department The University of Texas at Austin rjiang@utexas.edu Aryan Mokhtari ECE Department The University of Texas at Austin mokhtari@austin.utexas.edu
Pseudocode Yes Algorithm 1 Accelerated Quasi-Newton Proximal Extragradient Method
Open Source Code No The paper does not provide a specific link or explicit statement about the availability of the source code for the described methodology.
Open Datasets No We perform our numerical experiments on a synthetic dataset and the data generation process is described in Appendix F. In the first experiment of logistic regression, the dataset consists of n data points {(ai, yi)}n i=1... we generate {ai}n i=1 by adding noises and appending an extra dimension to {a i }n i=1.
Dataset Splits No The paper describes generating synthetic datasets but does not provide any specific information about training, validation, or test splits, nor does it refer to standard splits.
Hardware Specification Yes All experiments are conducted using MATLAB R2021b on a Mac Book Pro with an Apple M1 chip and 16GB RAM.
Software Dependencies Yes All experiments are conducted using MATLAB R2021b
Experiment Setup Yes In our experiment, we set n = 2,000, d = 150 and σ = 0.8. In our experiment, we set n = d = 250. We also use a line search scheme in NAG and BFGS to obtain their best performance.