Generalized Linear Bandits With Safety Constraints

The classical multi-armed bandit is a class of sequential decision making problems where selecting actions incurs costs that are sampled independently from an unknown underlying distribution. Bandit algorithms have many applications in safety critical sys