Online Learning with Improving Agents: Multiclass, Budgeted Agents and Bandit Learners
This work addresses incremental improvements in machine learning models where agents can adjust features for better outcomes, relevant for applications like algorithmic decision-making.
The paper tackles the problem of online learning with improving agents by extending previous results to multiclass classification, bandit feedback, and cost modeling, providing combinatorial dimensions that characterize online learnability in this model.
We investigate the recently introduced model of learning with improvements, where agents are allowed to make small changes to their feature values to be warranted a more desirable label. We extensively extend previously published results by providing combinatorial dimensions that characterize online learnability in this model, by analyzing the multiclass setup, learnability in a bandit feedback setup, modeling agents' cost for making improvements and more.