Bridging the Gap between General and Down-Closed Convex Sets in Submodular Maximization

Authors: Loay Mualem, Murad Tukan, Moran Feldman

IJCAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We also empirically demonstrate the superiority of our proposed algorithms across three offline and two online applications. Our offline and online algorithms...can be found in Sections 4 and 5, respectively. Finally, Section 6 compares the empirical performance of our algorithms on multiple machine learning applications with the performance of previously suggested algorithms from the literature.
Researcher Affiliation Collaboration Loay Mualem1 , Murad Tukan2 and Moran Feldman1 1University of Haifa 2Data Heroes loaymua@gmail.com, murad@dataheroes.ai, moranfe@cs.haifa.ac.il
Pseudocode Yes Algorithm 1 Frank-Wolfe/Continuous-Greedy Hybrid for Known F(o(1) D ) Algorithm 2 Online Frank-Wolfe/Continuous -Greedy Hybrid
Open Source Code No The paper does not provide any statement or link indicating that the source code for the methodology is openly available.
Open Datasets Yes The first dataset is sourced from a Facebook network [Viswanath et al., 2009], encompassing 64K users (vertices) and 1M unweighted relationships (edges). The second dataset is based on the Advogato network [Massa et al., 2009], comprising 6.5K users (vertices) and 61K weighted relationships (edges).
Dataset Splits No The paper mentions using specific datasets but does not provide explicit details on training, validation, or test splits. It describes setting a number of time steps for online experiments and specific graph sizes, but not data partitioning for model training and evaluation in a typical train/validation/test sense.
Hardware Specification No The paper does not provide any specific details about the hardware used to run the experiments (e.g., GPU models, CPU types, or memory).
Software Dependencies No The paper mentions that algorithms were implemented but does not provide specific software names with version numbers for dependencies (e.g., programming languages, libraries, or frameworks).
Experiment Setup Yes In both algorithms, we have set the number of online linear optimizers used to be 100 (which corresponding to setting the error parameter ε to 0.01 in our algorithm and to ln 2/100 in the algorithm of Mualem & Feldman). We set the number of time steps to L = 1000, with p = 0.0001. The above objective functions are optimized subject to the constraint 0.1 P i xi 1. Both algorithms have been executed for T = 100 iterations, and the error parameter ε was set accordingly.