Fairness of Exposure in Stochastic Bandits
Authors: Lequn Wang, Yiwei Bai, Wen Sun, Thorsten Joachims
ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Beyond the theoretical analysis, we also provide empirical evidence that these algorithms can fairly allocate exposure to different arms effectively. |
| Researcher Affiliation | Academia | 1Department of Computer Science, Cornell University, Ithaca, NY, USA. |
| Pseudocode | Yes | Algorithm 1 Fair X-UCB Algorithm; Algorithm 2 Fair X-TS Algorithm; Algorithm 3 Fair X-Lin UCB Algorithm; Algorithm 4 Fair X-Lin TS Algorithm |
| Open Source Code | No | The paper does not provide any statement or link indicating the release of source code for the described methodology. |
| Open Datasets | Yes | For the experiments where we control the properties of the synthetic data, we derive bandit problems from the multi-label datasets yeast (Horton & Nakai, 1996) and mediamill (Snoek et al., 2006). ... For the experiments on real-world data, we use data from the Yahoo! Today Module (Li et al., 2010) |
| Dataset Splits | Yes | We randomly split the dataset into two sets, 20% as the validation set to tune hyper-parameters and 80% as the test set to test the performance of different algorithms. ... We use the data from the first day for hyperparameter selection and report the results on the data from the second day. |
| Hardware Specification | No | No specific hardware details (e.g., GPU/CPU models, memory amounts) are mentioned for running experiments. |
| Software Dependencies | No | The paper does not provide specific version numbers for software dependencies or libraries used in the experiments. |
| Experiment Setup | Yes | We set the learning rate to be 0.01 and the number of steps to be 10. For Fair XLin UCB, we use a fixed βt = β for all rounds. ... We grid search w0 for Fair X-UCB and UCB; prior variance and reward variance for Fair X-TS, TS, Fair X-Lin TS and Lin TS; λ and β for Fair X-Lin UCB and Lin UCB; ϵ for Fair X-EG; ϵ and the regularization parameter of the ridge regression for Fair XLin EG. |