The field of online learning and decision-making is moving towards more sophisticated and adaptive methods. Researchers are exploring ways to improve the performance of online learning algorithms in complex and dynamic environments, such as those with endogenous uncertainty or biased offline data. One notable direction is the development of hierarchical and recursive algorithms that can learn and adapt to changing conditions. Another area of focus is the integration of machine learning models with decision-making processes, enabling more effective and robust decision-making under uncertainty. Noteworthy papers in this area include:

• A hierarchical Vovk-Azoury-Warmuth forecaster with discounting for online regression in RKHS, which achieves optimal dynamic regret in the non-parametric domain.
• Aligning Learning and Endogenous Decision-Making, which introduces an end-to-end method for training ML models to be aware of their downstream impact and enables effective use in decision-making.