Adaptive Submodular

Adaptive submodularity addresses optimization problems where the value of selecting an item depends on previously selected items, arising in diverse applications like sensor placement and active learning. Current research focuses on developing efficient algorithms for maximizing adaptive submodular functions under various constraints, including cardinality, knapsack, and group equality constraints, often employing greedy or random greedy approaches and exploring the impact of monotonicity. These advancements improve the efficiency and fairness of solutions for numerous machine learning and resource allocation problems, offering better approximation guarantees and handling streaming data or concept drift.