Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Fair Clustering Under a Bounded Cost
Authors: Seyed Esmaeili, Brian Brubach, Aravind Srinivasan, John Dickerson
NeurIPS 2021 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conclude with experimental results on real-world datasets that demonstrate the validity of our algorithms. We validate our algorithms on datasets from the UCI repository [24]. The results here are for k-means clustering; additional experiments are in Appendix F. |
| Researcher Affiliation | Academia | Seyed Esmaeili University of Maryland EMAIL Brian Brubach Wellesley College EMAIL Aravind Srinivasan University of Maryland EMAIL John P. Dickerson University of Maryland EMAIL |
| Pseudocode | Yes | Algorithm 1 :ALG-FCBC(U, UNFAIRNESS-OBJECTIVE) and Algorithm 2 :ALG-FABC(S, U, UNFAIRNESS-OBJECTIVE) |
| Open Source Code | No | The paper mentions using third-party libraries like Scikit-learn and Network X but does not state that the authors' own implementation code for the described methodology is publicly available. |
| Open Datasets | Yes | We use all 32,561 entries of the Adult dataset [34]. For the Census1990 dataset [41], because of its large size (over 2 million points) we sub-sample the dataset to a range similar to that considered in the fair clustering literature [20, 10]; specifically we use 20,000 data points. We also use the Credit Card dataset [47] which has 30,000 points (results are in Appendix F). |
| Dataset Splits | No | The paper mentions the datasets used but does not specify any training, validation, or test split percentages or sample counts, nor does it refer to predefined splits. |
| Hardware Specification | No | The paper states 'We only use commodity hardware for all experiments' but does not provide specific details such as CPU/GPU models or memory amounts. |
| Software Dependencies | Yes | We only use commodity hardware for all experiments with our programs running on Python 3.6. ... Our LPs are solved using CPLEX [32]. Scikit-learn [46] is called for subroutines such as k-means++. The network-flow rounding is handled using Network X [25]. |
| Experiment Setup | Yes | We set the upper and lower bounds for each color to αh = (1+δ)rh and βh = (1 δ)rh. ... Further, for all experiments we discretize the space by ϵ = 1 27 < 0.008. ... We set δ = 0.05 and k = 10 for Adult and δ = 0.1 and k = 5 for Census1990. |