Achieving Long-Term Fairness in Submodular Maximization Through Randomization
摘要
Submodular function optimization is applied in ML and data analysis, including diverse dataset summarization. Fairness-aware algorithms are essential for handling sensitive attributes. Our research investigates the problem of maximizing a monotone submodular function while adhering to constraints on the expected number of selected items per group. Our goal is to compute a distribution over feasible sets, and to achieve this, we develop a series of approximation algorithms.