Learning-Augmented Algorithms for the Bahncard Problem

Authors: Hailiang Zhao, Xueyan Tang, Peng Chen, Shuiguang Deng

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

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct extensive experiments to compare PFSUM with SUM [1], PDLA [12], SUMw (Section 4.1), FSUM (Section 4.2), and SRL (Ski-Rental-based Learning algorithm), where SRL is adapted from Algorithm 2 proposed in [4] for ski-rental.
Researcher Affiliation Academia Hailiang Zhao1 Xueyan Tang2 Peng Chen1 Shuiguang Deng1 1Zhejiang University 2Nanyang Technological University {hliangzhao,pgchen,dengsg}@zju.edu.cn asxytang@ntu.edu.sg
Pseudocode No The paper describes algorithms (SUM, FSUM, PFSUM) in narrative text within the main body, but does not include any structured pseudocode or algorithm blocks.
Open Source Code Yes We have released the code and data for reproducibility. With our instructions, others can easily reproduce experimental results that are consistent with the results in the paper. (NeurIPS Checklist, Question 5)
Open Datasets Yes Input instances. Referring to the experimental setup of [31], we consider two main types of traveler profiles: commuters, and occasional travelers. We set the timespan at 2000 days... For each request sequence generated, we investigate three types of ticket price distributions: a Normal distribution... a Uniform distribution... and a Pareto distribution... The code and data can be openly accessed. (NeurIPS Checklist, Question 5)
Dataset Splits No The paper describes generating multiple input instances and running online algorithms on them; however, it does not specify explicit training, validation, or test dataset splits for model training or evaluation, as the algorithms are not traditional machine learning models requiring such splits.
Hardware Specification No The paper does not explicitly provide details about the specific hardware (e.g., GPU/CPU models, memory) used for running the experiments within the main text or appendices.
Software Dependencies No The paper does not explicitly list specific software dependencies with their version numbers (e.g., Python, PyTorch, CUDA versions) required to replicate the experiments.
Experiment Setup Yes We test SUMw with w = T/2, and test SRL and PDLA with λ set to 0.2, 0.5, and 1. To accommodate SRL and PDLA, we discretize time over a sufficiently long timespan, closely approximating a continuous time scenario.