Data-driven Competitive Algorithms for Online Knapsack and Set Cover
Authors: Ali Zeynali, Bo Sun, Mohammad Hajiesmaili, Adam Wierman10833-10841
AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Additionally, we illustrate the practical implications via a case study on electric vehicle charging. Finally, to demonstrate the potential for achieving both data-driven adaptation and worst-case guarantees in practical settings, we apply our approach to the application of online admission control of electric vehicles in a charging station... Experimental setup. To explore the performance of our framework for EV admission control, we use ACNData (Lee, Li, and Low 2019), an open EV dataset... |
| Researcher Affiliation | Academia | 1 University of Massachusetts Amherst, USA 2 The Hong Kong University of Science and Technology, Hong Kong 3 California Institute of Technology, USA |
| Pseudocode | Yes | Algorithm 1 OKP-Alg(α): A parametric algorithm for OKP; Algorithm 2 OSC-Alg(θ): A parametric algorithm for OSC |
| Open Source Code | No | The proofs are given in (Zeynali et al. 2020). |
| Open Datasets | Yes | To explore the performance of our framework for EV admission control, we use ACNData (Lee, Li, and Low 2019), an open EV dataset including over 50,000 fine-grained EV charging sessions. |
| Dataset Splits | No | In the experiments, each instance considers one day, and we randomly generated 100 instances for each day, each with different values, and report the results for 60 × 100 = 6000 instances. |
| Hardware Specification | No | No specific hardware details (GPU/CPU models, processor types, or memory amounts) used for running the experiments were provided in the paper. |
| Software Dependencies | No | No specific ancillary software details (library or solver names with version numbers) needed to replicate the experiment were provided in the paper. |
| Experiment Setup | No | The paper mentions using a value estimation approach and a water-filling scheduling policy, and an adversarial Lipschitz learning algorithm, but does not provide specific hyperparameter values or detailed system-level training settings for these components. |