Mitigating Label Bias in Machine Learning: Fairness through Confident Learning

Authors: Yixuan Zhang, Boyu Li, Zenan Ling, Feng Zhou

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Through extensive experimentation and evaluation of various datasets, we demonstrate the efficacy of our approach in promoting fairness and reducing the impact of label bias in machine learning models. In this section, we demonstrate the effectiveness of our methods by comparing them with several baseline models on the benchmark datasets.
Researcher Affiliation Academia Yixuan Zhang1, Boyu Li2, Zenan Ling3, Feng Zhou4* 1China-Austria Belt and Road Joint Laboratory on AI and AM, Hangzhou Dianzi University, China 2Data Science Institute, University of Technology Sydney, Australia 3School of Electronic Information and Communications, Huazhong University of Science and Technology, China 4Center for Applied Statistics and School of Statistics, Renmin University of China, China
Pseudocode Yes Algorithm 1: Training Algorithm
Open Source Code No No explicit statement regarding the release of open-source code or a link to a code repository is provided.
Open Datasets Yes Adult (Kohavi 1996), Pro Publica COMPAS (Brennan, Dieterich, and Ehret 2009), Credit Loan Data (Yeh 2016), Law School Admissions (Wightman 1998).
Dataset Splits Yes To ensure robust results, we perform 10 rounds of random shuffling on the training set, while retaining 10% of the biased training examples as a validation set for hyperparameter optimization.
Hardware Specification No The paper does not provide specific details about the hardware used for running experiments.
Software Dependencies No The paper does not list specific software dependencies with version numbers.
Experiment Setup Yes Consistent hyperparameters are maintained across all experiments for all methods. Additional implementation details can be found in the Appendix. We explore the impact of hyperparameters ν in the range of {10^-4, 10^-3, 10^-2, 10^-1} and Ns in the range of {0.5, 0.6, 0.7, 0.8, 0.9} to examine their impact using synthetic data. We label the methods using Eq. (3) as M with hyperparameters Ns = 0.6 and ν = 10^-2 fixed and we label the method using Eq. (5) as T.