Cluster Randomized Designs for One-Sided Bipartite Experiments
Authors: Jennifer Brennan, Vahab Mirrokni, Jean Pouget-Abadie
NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 5 Experiments We explore the performance of our cluster-based randomized design in several settings using simulated graphs. |
| Researcher Affiliation | Collaboration | Jennifer Brennan Paul G. Allen School of Computer Science & Engineering, University of Washington Seattle, WA 98195 jrbrennan@google.com Vahab Mirrokni Google Research New York, NY 10011 mirrokni@google.com Jean Pouget-Abadie Google Research New York, NY 10011 jeanpa@google.com |
| Pseudocode | No | The paper does not contain any figures or sections explicitly labeled 'Pseudocode' or 'Algorithm'. |
| Open Source Code | No | The balanced partitioning code was shared with us by the authors of Aydin et al. [35], and code to optimize the EXPOSURE-DESIGN objective was shared by the authors of Harshaw et al. [19], but neither is publicly available. |
| Open Datasets | No | All data was simulated. and We construct a synthetic graph according to the bipartite stochastic block model with N = 1, 000 experimental units and M = 2, 000 interference units. |
| Dataset Splits | No | The paper describes experiments based on simulated graphs and random draws of treatment assignments, but it does not specify explicit train, validation, or test dataset splits in terms of percentages or counts for reproducibility. |
| Hardware Specification | No | The paper does not provide any specific hardware details such as GPU models, CPU types, or memory specifications used for running the experiments. |
| Software Dependencies | No | The paper mentions using code provided by the authors of [35] and [19], and extending code from [10], but it does not list specific software dependencies with version numbers (e.g., 'Python 3.8, PyTorch 1.9'). |
| Experiment Setup | Yes | We construct a synthetic graph according to the bipartite stochastic block model with N = 1, 000 experimental units and M = 2, 000 interference units. Both sides of the graph are partitioned into 20 equally sized groups with label i [20]... All clustering designs used K = 20 clusters, with KT = 10 clusters assigned to treatment. |