Adaptive Sample Sharing for Multi Agent Linear Bandits

Authors: Hamza Cherkaoui, Merwan Barlier, Igor Colin

ICML 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our main result formalizes the trade-off between the bias and uncertainty of the bandit parameter estimation for efficient collaboration. This result is the cornerstone of the Bandit Adaptive Sample Sharing (BASS) algorithm, whose efficiency over the current state-of-the-art is validated through both theoretical analysis and empirical evaluations on both synthetic and real-world datasets.
Researcher Affiliation Collaboration 1Huawei Noah s Ark Lab, France 2LTCI, Télécom Paris, Institut Polytechnique de Paris, France. Correspondence to: Hamza Cherkaoui <EMAIL>.
Pseudocode Yes Algorithm 1 Bandit Adaptive Sample Sharing (BASS) algorithm
Open Source Code Yes The code is publicly available and can be found at this repository.
Open Datasets Yes We consider two public datasets of ranking scenarios: Movie Lens and Yahoo! dataset.
Dataset Splits No The paper describes synthetic data generation and preprocessing for real datasets (Movie Lens and Yahoo!) but does not specify explicit training, validation, or test dataset splits in terms of percentages, sample counts, or predefined split references for reproducibility.
Hardware Specification Yes The experiments were run in Python on 50 Intel Xeon @ 3.20 GHz CPUs and lasted a week.
Software Dependencies No The paper mentions that experiments were run in Python, but it does not provide specific version numbers for Python itself or any other software libraries or dependencies used (e.g., PyTorch, TensorFlow, scikit-learn).
Experiment Setup Yes For each algorithm, we line-search the α parameter. We set our hyperparameter γ to 2 to match Gilitschenski & Hanebeck (2012). For the BASS algorithm, we test two values for δ {0.1, 0.9}.