Cascading Bandits: Learning to Rank in the Cascade Model
This work addresses the problem of optimizing search engine rankings for users by efficiently learning from feedback in a cascade model, representing an incremental improvement with theoretical guarantees.
The paper tackles the problem of learning to rank items in a cascade model, where a user sequentially examines a list and selects the first attractive item, by proposing cascading bandits algorithms to identify the top K items. The result includes algorithms like CascadeUCB1 and CascadeKL-UCB with proven gap-dependent regret bounds, where the lower bound matches the upper bound up to a logarithmic factor, and experiments show strong performance even under violated assumptions.
A search engine usually outputs a list of $K$ web pages. The user examines this list, from the first web page to the last, and chooses the first attractive page. This model of user behavior is known as the cascade model. In this paper, we propose cascading bandits, a learning variant of the cascade model where the objective is to identify $K$ most attractive items. We formulate our problem as a stochastic combinatorial partial monitoring problem. We propose two algorithms for solving it, CascadeUCB1 and CascadeKL-UCB. We also prove gap-dependent upper bounds on the regret of these algorithms and derive a lower bound on the regret in cascading bandits. The lower bound matches the upper bound of CascadeKL-UCB up to a logarithmic factor. We experiment with our algorithms on several problems. The algorithms perform surprisingly well even when our modeling assumptions are violated.