Improved Algorithm for Regret Ratio Minimization in Multi-Objective Submodular Maximization

Authors: Yanhao Wang, Jiping Zheng, Fanxu Meng

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on real-world and synthetic data confirm that HS-RRM achieves lower MRRs than existing algorithms.In this section, we experimentally compare HS-RRM with four existing algorithms for RRM: COORDINATE and POLYTOPE in (Soma and Yoshida 2017), and RRMS and RRMS in (Feng and Qian 2021).
Researcher Affiliation Academia 1School of Data Science and Engineering, East China Normal University, Shanghai, China 2College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing, China 3State Key Laboratory for Novel Software Technology, Nanjing University, Nanjing, China
Pseudocode Yes Algorithm 1: HS-RRM
Open Source Code Yes Our code and data are publicly available at https://github.com/yhwang1990/code-rrm-release.
Open Datasets Yes We use a real-world dataset email-Eu-core from SNAP1, which is a directed graph with 1,004 vertices and 25,571 edges... and 1http://snap.stanford.edu/data/#email. In our experiments, we use the Movie Lens dataset2 that contains 100,000 ratings from 943 users on 1,682 movies. and 2https://grouplens.org/datasets/movielens/
Dataset Splits No No explicit train/validation/test dataset splits were provided for the datasets used (email-Eu-core, Movie Lens). The paper mentions 'sampling a validation set of 1,000 functions' but this is for MRR estimation, not for data splitting.
Hardware Specification Yes Our experiments were conducted on a server with an Intel Xeon W-2123 3.60GHz processor and 32GB memory running Ubuntu 18.04.
Software Dependencies No The paper states 'All the algorithms were implemented in Python 3.' but does not list specific versions for any libraries or dependencies.
Experiment Setup Yes We set the dimensionality of user and movie vectors to 25. We run the k-means clustering (Lloyd 1982; Arthur and Vassilvitskii 2007) to partition the movies into d subsets. Like WMC, the feasible solution set I is defined by r = 10, and each algorithm runs ten times independently. ... In a few cases (e.g., k ranges from 8 to 10 when d = 2), the MRRs of the output of HS-RRM are slightly higher than those of POLYTOPE and RRMS. This is because we set λ = 10 3 for HS-RRM when d = 2.