Reinforcement Learning and Regret Bounds for Admission Control

Authors: Lucas Weber, Ana Busic, Jiamin Zhu

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental evaluations are conducted in section 7
Researcher Affiliation Collaboration 1Inria and DI ENS, Ecole Normale Sup erieure, PSL University, Paris, France 2IFP Energies nouvelles, 1 et 4 avenue de Bois-Pr eau, 92852 Rueil-Malmaison, France.
Pseudocode Yes Algorithm 1 PI for Admission Control; Algorithm 2 VI for Admission Control; Algorithm 3 UCRL-AC
Open Source Code Yes Our code is available at: https://github.com/luweber21/ucrl-ac.
Open Datasets No The paper describes a simulated queuing system (M/M/c/S queue) and its parameters rather than using a conventional publicly available dataset.
Dataset Splits No The paper simulates a queuing system and does not specify train/validation/test splits as it does not rely on a conventional dataset.
Hardware Specification Yes The experiments were run on a Mac Book Pro 2021 equipped with an Apple M1 Pro processor and 16 GB RAM.
Software Dependencies No The paper does not specify version numbers for key software components or libraries used in the experiments.
Experiment Setup Yes We consider 5 servers, 2 job classes with immediate rewards R1 = 20 and R2 = 10 and arrival rates λ1 = 1 and λ2 = 1 respectively, and holding cost C(t) = 0.1t for both classes. For UCRL-AC, we used Λmin = 1 and Λmax = 4.