This assignment is part of the CO557 Ethics, Privacy, AI in Society module of Imperial College London. The task is the analysis of the effect of regularisation on accuracy-fairness trade-off obtained by a binary classifier on a dataset with sensitive attributes.
-
Code:
Adult Dataset.ipynb
contains an analysis on the Adult Dataset.Compas Dataset.ipynb
contains an analysis on the Compas Dataset.Adult non-binary.ipynb
contains an analysis on the Adult Dataset in the case of a non-binary sensitive feature as an extrapolation of what was taught.Model_Selection.ipynb
contains the random train/splits and average results for model selection.
-
Report