Abstract:
Machine learning’s fairness is a growing concern. Most machine-learning research is focused on either supervised or unsupervised learning. Two observations led us to consider semi-supervised learning for discrimination problems. First, previous research suggested that a larger training set may improve fairness-accuracy trade-offs. Second, the most powerful models today require massive amounts of data to train, which is likely possible from a mix of labeled and unlabeled data. Thus, this paper presents a framework for fair semi-supervised learning in the pre-processing phase, including pseudo labeling to predict labels for unlabeled data, re-sampling to obtain multiple fair datasets, and ensemble learning to improve accuracy and decrease discrimination. A theoretical decomposition analysis of bias, variance, and noise shows how different sources of discrimination affect semi-supervised learning fairness. Our method can use unlabeled data to improve accuracy and discrimination, according to real-world and synthetic dataset experiments.
Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.
Did you like this final year project?
To download this project Code with thesis report and project training... Click Here
