Bayesian Optimization-Based Combinatorial Assignment

We study the combinatorial assignment domain, which includes auctions and course allocation. The main challenge in this domain is that bundle space grows exponentially number of items. To address this, several papers have recently proposed machine learning-based preference elicitation algorithms aim to elicit only most important information from agents. However, shortcoming prior work it does not model a mechanism's uncertainty over values for yet elicited bundles. In paper, we by presenting Bayesian optimization-based (BOCA) mechanism. Our key technical contribution integrate method capturing into an iterative auction Concretely, design new estimating upper bound can be used define acquisition function determine next query This enables mechanism properly explore (and just exploit) during its phase. run computational experiments spectrum domains evaluate BOCA's performance. results show BOCA achieves higher allocative efficiency than state-of-the-art approaches.