Source Code and Curated Datasets for "Hypervolume-Contribution Bandits for Multi-Objective Feature Selection with Nested Cross-Validation

Read full text Summarise with VIRAT
Authors: Jing Li, Jing Li
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This repository provides the source code and curated datasets for the paper:“Hypervolume-Contribution Bandits for Multi-Objective Feature Selection with Nested Cross-Validation”. OverviewWe propose RL-HMOFS, a lightweight contextual-bandit framework for multi-objective feature selection (MOFS) in high-dimensional, small-sample settings. RL-HMOFS adaptively schedules a pool of variation operators, using a reward based on marginal Hypervolume Contribution (HVC) computed on the selected Pareto front. Reproducible evaluation (Nested CV)All experiments follow a strict nested cross-validation protocol to avoid selection bias: the MOEA search and model selection are performed only on outer-training folds, while all reported objective values and performance indicators are evaluated exclusively on outer-test folds. Contents Source code: RL-HMOFS, Random-HMOFS (random operator scheduling ablation), and the NSGA-II baseline; including the operator pool (Standard, MI-guided, Reduce) and the nested-CV evaluation pipeline. Datasets & Attribution: 28 pre-processed benchmark datasets used in the paper. The original raw data were sourced from public repositories, including the UCI Machine Learning Repository, OpenML, and the ASU Feature Selection Lab. To ensure 100% exact reproducibility of the experimental results in our paper, we provide our pre-processed CSV versions. Specifically, non-numeric columns were removed, and original multi-class datasets were formatted into binary classification tasks by selecting the top two most frequent classes. Users are encouraged to cite the original data sources when utilizing these datasets for other purposes. Citation Jing Li, et al. "Hypervolume-Contribution Bandits for Multi-Objective Feature Selection with Nested Cross-Validation", Submitted2026.]

Fields of Study & Subjects

Advanced Multi-Objective Optimization Algorithms Evolutionary Algorithms and Applications Advanced Bandit Algorithms Research

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
28 March 2026
Language
EN