The coursework for this sub-module required the creation of a machine learning model which mitigated bias towards a single protected class. For my project, I decided to mitigate bias between genders on the UCI Adult Income dataset, using the Classification with No Discrimination method outlined by Kamiran and Calders.
The dataset used in this repository was downloaded from datahub and is derived from the original UCI dataset.
The project was expected to be delivered alongside a 3 page report which outlined the purpose of the project and assessed the effects of the chosen mitigation method.
The report can be found in project_report.pdf.
It was also required to submit a 1-page essay which reviewed a selected paper on the topic of Bias in AI and presented my own views on the future of the field.
Full feedback for the assignment can be found in feedback.txt. The final mark received was 97.5%.
By boyla950.