Multi-objective Bandits: Optimizing the Generalized Gini Index
Authors: Róbert Busa-Fekete, Balázs Szörényi, Paul Weng, Shie Mannor
ICML 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We test our algorithm on synthetic data as well as on an electric battery control problem where the goal is to trade off the use of the different cells of a battery in order to balance their respective degradation rates. |
| Researcher Affiliation | Collaboration | 1Yahoo Research, New York, NY, USA 2Research Group on AI, Hungarian Acad. Sci. and Univ. of Szeged, Szeged, Hungary 3Technion Institute of Technology, Haifa, Israel 4SYSU-CMU JIE, SEIT, SYSU, Guangzhou, P.R. China 5SYSU-CMU JRI, Shunde, P.R. China. |
| Pseudocode | Yes | Algorithm 1 MO-OGDE(δ) |
| Open Source Code | No | The paper does not provide any explicit statement or link indicating that the source code for the described methodology is publicly available. |
| Open Datasets | No | The synthetic data was generated for the experiments. For the battery control task, the paper states: 'Data are provided by our industry partner, and will be made publicly available soon.' This indicates it is not currently publicly available. |
| Dataset Splits | No | The paper does not explicitly provide details about training, validation, and test dataset splits. The experiments are conducted in an online learning setting, where data is revealed sequentially, and performance is measured by regret over rounds. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., CPU, GPU models, memory) used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific software names with version numbers that would be necessary to replicate the experiments. |
| Experiment Setup | Yes | We ran the MOOGDE and MO-LP algorithms with 100 repetitions. The number of arms K was set to {5, 20} and the dimension of the cost distribution was taken from D 2 {5, 10}. The weight vector w of GGI was set to wd = 1/2d 1. |