Improving Policy-Constrained Kidney Exchange via Pre-Screening
Authors: Duncan McElfresh, Michael Curry, Tuomas Sandholm, John Dickerson
NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct numerical experiments on both simulated and real exchange data from the United Network for Organ Sharing (UNOS). |
| Researcher Affiliation | Collaboration | Duncan C Mc Elfresh Mathematics Department Computer Science Department University of Maryland College Park, MD 20742 dmcelfre@umd.edu Michael Curry Computer Science Department University of Maryland College Park, MD 20742 curry@cs.umd.edu Tuomas Sandholm Computer Science Department Carnegie Mellon University Strategy Robot, Inc. Optimized Markets, Inc. Strategic Machine, Inc. John P Dickerson Computer Science Department University of Maryland College Park, MD 20742 john@cs.umd.edu |
| Pseudocode | Yes | ALGORITHM 1: MCTS: Tree Search for Single-Stage Edge Selection ALGORITHM 2: Sample: Sampling function used by MCTS |
| Open Source Code | Yes | all code for these experiments is available online.4 4https://github.com/duncanmcelfresh/kpd-edge-query |
| Open Datasets | No | We use exchange graphs from the United Network for Organ Sharing (UNOS), representing UNOS match runs between 2010 and 2019. ... UNOS is the organization tasked with overseeing organ transplantation in the US: https://unos.org/. While the paper mentions data from UNOS, it provides a link to the organization's general website, not a direct link, DOI, or specific citation for the dataset used in the experiments. The synthetic data generation is described, but it's not a publicly available dataset. |
| Dataset Splits | No | The paper describes the data used (UNOS and synthetic) and how models are evaluated, but it does not specify any explicit training, validation, or test dataset splits (e.g., percentages or sample counts). |
| Hardware Specification | No | The paper does not specify any particular hardware (e.g., GPU models, CPU types, or cloud computing instances with their specifications) used for running the experiments. |
| Software Dependencies | No | The paper mentions that code is available online but does not specify any particular software dependencies or their version numbers (e.g., Python, PyTorch, or specific solvers like CPLEX) required to replicate the experiments. |
| Experiment Setup | Yes | For MCTS we use a 1-hour time limit per edge (Γ hours total). ... MCTS uses a 1-hour training time per level. ... We generate three sets of 100 random graphs with N = 50, 75, and 100 vertices, and each with p = 0.01. ... In the Simple distribution, PR = 0.5, PQ = 1, and PN = 0.5 for all edges. The KPD distribution is inspired by the fielded exchange setting from which we draw our real underlying compatibility graphs. ... we draw PR uniformly from U(0.25, 0.43) for each edge. ... for these edges we draw PQ from U(0.2, 0.5) and PN from U(0.0, 0.2). For other edges we draw PQ from U(0.9, 1.0) and PN from U(0.8, 0.9). |