Primal-Dual Rates and Certificates

Authors: Celestine Dünner, Simone Forte, Martin Takac, Martin Jaggi

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

Reproducibility Variable Result LLM Response
Research Type Experimental Here we illustrate the usefulness of our framework by showcasing it for two important applications, each one showing two algorithm examples for optimizing (A). The top row of Figure 2 shows the primal-dual convergence of Algorithm 1 (CD) as well as the accelerated variant of CD (APPROX, Fercoq & Richt arik (2015)), both applied to the Lasso problem (A). On the bottom row of Figure 2 we compare CD with its accelerated variant on two benchmark datasets.5 We have chosen λ = 1/n.
Researcher Affiliation Collaboration Celestine D unner CDU@ZURICH.IBM.COM IBM Research, Z urich, Switzerland Simone Forte FORTESIMONE90@GMAIL.COM ETH Z urich, Switzerland Martin Tak aˇc TAKAC.MT@GMAIL.COM Lehigh University, USA Martin Jaggi JAGGIM@INF.ETHZ.CH ETH Z urich, Switzerland
Pseudocode Yes Algorithm 1 Coordinate Descent on D(α)
Open Source Code No The paper does not provide any concrete access to source code for the methodology described.
Open Datasets Yes On the bottom row of Figure 2 we compare CD with its accelerated variant on two benchmark datasets.5 We have chosen λ = 1/n. Footnote 5: Available from csie.ntu.edu.tw/ cjlin/libsvmtools/datasets.
Dataset Splits No The paper mentions using benchmark datasets but does not specify exact training/validation/test split percentages, sample counts, or explicitly reference predefined splits with citations.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types, or memory amounts) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers) needed to replicate the experiments.
Experiment Setup Yes On the bottom row of Figure 2 we compare CD with its accelerated variant on two benchmark datasets.5 We have chosen λ = 1/n.