Strategy-Proof and Efficient Kidney Exchange Using a Credit Mechanism
Authors: Chen Hajaj, John Dickerson, Avinatan Hassidim, Tuomas Sandholm, David Sarne
AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we evaluate the mechanism experimentally. We use data from a large, fielded kidney exchange in the US run by the United Network for Organ Sharing (UNOS). The exchange started in 2010 and now includes 3-cycles and 4-chains, in accordance with the current practice of the UNOS exchange. We also include experiments on an older family of graphs in kidney exchange research, from a generator due to Saidman et al. (2006). |
| Researcher Affiliation | Academia | Chen Hajaj Bar-Ilan University chen.hajaj@biu.ac.il John P. Dickerson Carnegie Mellon University dickerson@cs.cmu.edu Avinatan Hassidim Bar-Ilan University avinatan@cs.biu.ac.il Tuomas Sandholm Carnegie Mellon University sandholm@cs.cmu.edu David Sarne Bar-Ilan University sarned@cs.biu.ac.il |
| Pseudocode | Yes | Algorithm 1 Credit-based matching mechanism. |
| Open Source Code | Yes | Our simulation framework was built in Java 1.6 on top of a vetted open source kidney exchange software suite;8 we incorporated our mechanisms into this suite. (Footnote 8: https://github.com/John Dickerson/Kidney Exchange) |
| Open Datasets | Yes | We use data from a large, fielded kidney exchange in the US run by the United Network for Organ Sharing (UNOS). The exchange started in 2010 and now includes over 140 transplant centers. ... We also include experiments on an older family of graphs in kidney exchange research, from a generator due to Saidman et al. (2006). |
| Dataset Splits | No | The paper describes its simulation setup and data generation but does not provide specific train/validation/test dataset splits (percentages or counts) or reference predefined splits. |
| Hardware Specification | No | The paper mentions using "Blacklight at the Pittsburgh Supercomputing Center (PSC)" but does not provide specific hardware details such as GPU/CPU models, processor types, or memory amounts. |
| Software Dependencies | No | The paper states it was "built in Java 1.6" and uses "IBM CPLEX as the IP solver", but it does not provide a version number for CPLEX or other specific ancillary software with versions. |
| Experiment Setup | Yes | At a high level, a single run of simulation executes as follows. First, fix the number of transplant centers |T|, arrival distribution and time limit η. Then, for each time period t {1, . . . , η}, we execute the mechanism as described in Algorithm 1. ... Arrival rates varied from very low (e.g., U[1, 5]... to very high (e.g., U[25, 35]... The percentage of altruists received was determined endogenously by the UNOS data, or set to 5% of the number of pairs in the Saidman case. ...Runs for each parameter setting were performed at least 25 times. |