Achieving Counterfactual Fairness for Causal Bandit

Authors: Wen Huang, Lu Zhang, Xintao Wu6952-6959

AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Both theoretical analysis and empirical evaluation demonstrate effectiveness of our algorithms. We conduct experiments on the Email Campaign data (Lu et al. 2020) whose results show the benefit of using the dseparation set from the causal graph.
Researcher Affiliation Academia Wen Huang, Lu Zhang, Xintao Wu University of Arkansas {wenhuang, lz006, xintaowu}@uark.edu
Pseudocode Yes Algorithm 1: D-UCB: Causal Bandit based on d-separation
Open Source Code No The paper does not contain an explicit statement about releasing source code or a link to a code repository for the methodology described.
Open Datasets Yes We adopt the Email Campaign data as used in previous works (Lu et al. 2020). We follow the settings of (Huang et al. 2020) by combining two publicly available datasets: Adult dataset and Youtube video dataset.
Dataset Splits No The paper describes using 'half of the data as the offline data to construct causal graph and adopt the other half to be user sequence and arm candidates for online recommendation' but does not specify explicit training, validation, and test splits with percentages or sample counts for model evaluation, nor does it mention cross-validation.
Hardware Specification No The paper does not provide specific details about the hardware used for running the experiments, such as GPU models, CPU types, or memory specifications.
Software Dependencies No The paper does not provide specific software dependency details, such as library names with version numbers (e.g., 'Python 3.8', 'PyTorch 1.9').
Experiment Setup Yes In our experiment, we set δ = 1/t2 for each t [T]. The cumulative regret is added up through 5000 rounds.