wangqianwen0418
commented
5 years ago @TonyXuQAQ
A possible logic for "machine learning discrimination":
a) an important problem, a growing body of literature;
b) introduce previous work: discrimination discovery at different stages (data, model)
c) nice work, but they are binary classifier: whether discrimination exists or not => hard to compare and evaluate
d) recently, some work aimts to quantify discrimiation. "A Unified Approach to Quantifying Algorithmic Unfairness: Measuring Individual &Group Unfairness via Inequality Indices". But still miss details to guide the users of machine learning model
e) introduce our work
:pencil: related work(1.18)
The idea evaluation slides include some references for machine learning discrimination. slides