Approximation Guarantees of Local Search Algorithms via Localizability of Set Functions

Authors: Kaito Fujii

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct experiments in sparse regression and structure learning of graphical models to confirm the prac tical efficiency of the proposed local search algorithms.
Researcher Affiliation Academia 1National Institute of Informatics, Tokyo, Japan. Correspon dence to: Kaito Fujii <fujiik@nii.ac.jp>.
Pseudocode Yes Algorithm 1 Local search algorithms for a matroid con straint. ... Algorithm 2 Local search algorithms for a p-matroid inter section or p-exchange system constraint (p 2).
Open Source Code No The paper does not contain an explicit statement or link providing concrete access to the source code for the methodology described.
Open Datasets No The paper mentions generating synthetic datasets (e.g., "We generate synthetic datasets with a partition matroid constraint.") but does not provide concrete access information such as a link, DOI, or a citation to a publicly available dataset.
Dataset Splits No The paper does not provide specific dataset split information (e.g., percentages or sample counts for training, validation, and test sets) to reproduce the data partitioning.
Hardware Specification Yes We conduct the experiments in a machine with Intel Xeon E3 1225 V2 (3.20 GHz and 4 cores) and 16 GB RAM.
Software Dependencies No All the algorithms are implemented in Python 3.6. ... We use the L-BFGS-G solver in scipy.optimize li brary for evaluating the value of f. ... max-weight matching problem solver in Netwerk X library. The paper specifies Python 3.6, but does not provide version numbers for the other mentioned libraries (scipy.optimize, Netwerk X).
Experiment Setup Yes We set (n, d, nc, np) = (200, 50, 5, 5) in one setting and (n, d, nc, np) = (1000, 100, 10, 10) in the other setting. ... We apply these methods to a partition matroid constraint with capacity n0 p for each n0 2 {1, , 10}. ... We set (|V |, d) = (10, 5) in one setting and (|V |, d) = (20, 7) in the other setting. ... We apply these methods to pseudo-log-likelihood maximization under a b-matching constraint for each b 2 {1, , d}.