Noise-tolerant fair classification

Authors: Alex Lamy, Ziyuan Zhong, Aditya K. Menon, Nakul Verma

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We finally show that our procedure is empirically effective on two case-studies involving sensitive feature censoring. and We demonstrate that it is viable to learn fair classifiers given noisy sensitive features.
Researcher Affiliation Collaboration Alexandre Lamy Columbia University a.lamy@columbia.edu Ziyuan Zhong Columbia University ziyuan.zhong@columbia.edu Aditya Krishna Menon Google adityakmenon@google.com Nakul Verma Columbia University verma@cs.columbia.edu
Pseudocode Yes Algorithm 1 Reduction-based algorithm for fair classification given noisy A. Input: Training set S = {(xi, yi, ai)}n i=1, scorer class F, fairness tolerance 0, fairness constraint ( ), fair classification algorithm Fair Alg, noise estimation algorithm Noise Est Output: Fair classifier f 2 F 1: ˆ , ˆβ Noise Est(S) 2: 0 (1 ˆ ˆβ) 3: return Fair Alg(S, F, , 0)
Open Source Code Yes Source code is available at https://github.com/AIasd/noise_fairlearn.
Open Datasets Yes We look at COMPAS, a dataset from Propublica (Angwin et al., 2016) and We consider the dataset law school, which is a subset of the original dataset from LSAC (Wightman, 1998).
Dataset Splits No The paper specifies a random 80-20 training-testing split for its experiments (e.g., Section 5.3, 5.4) but does not explicitly mention a separate validation split.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., CPU, GPU models, memory, or cloud instances) used for running the experiments.
Software Dependencies No The paper does not specify the versions of any key software components or libraries used for the implementation or experiments.
Experiment Setup No While the paper details the noise settings for the datasets (e.g., CCN noise with + = = 0.15 is added to the sensitive attribute), it does not explicitly provide specific hyperparameters or system-level training settings for the classification model, such as learning rates, batch sizes, or optimizer configurations.