Improved Policy Evaluation for Randomized Trials of Algorithmic Resource Allocation

Authors: Aditya Mate, Bryan Wilder, Aparna Taneja, Milind Tambe

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate the value of our approach through empirical experiments on synthetic, semisynthetic as well as real case study data and show improved estimation accuracy across the board.
Researcher Affiliation Collaboration 1School of Engineering and Applied Science, Harvard University, USA 2Google Research, India 3Carnegie Mellon University, USA.
Pseudocode Yes Algorithm 1 Estimation through Assignment Permutation, Algorithm 2 Reshuffling between index based policies
Open Source Code No The paper does not provide any explicit statements about releasing source code or links to a code repository.
Open Datasets Yes We use real tuberculosis medication adherence monitoring data, consisting of daily records of patients in Mumbai, India, obtained from (Killian et al., 2019) and by considering an actual large-scale RCT reported in (Mate et al., 2022) evaluating a Restless Multi-Armed Bandit-based algorithm for resource allocation in a maternal and child healthcare.
Dataset Splits No The paper describes simulating data and running trial instances but does not provide specific train/validation/test dataset splits, percentages, or sample counts for model training and evaluation.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory, or cloud instance types) used for running the experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers for reproducibility.
Experiment Setup Yes We simulate N = 1000 patients in each arm and vary the budget constraint. We consider both the multi-step setting (T = 10) and single step (T = 1). We simulate 300 P1 individuals, 300 P2 individuals and (300 η P3) individuals and set an intervention budget of 300 per timestep for T = 20 timesteps.