Host: The Japanese Society for Artificial Intelligence
Name : The 33rd Annual Conference of the Japanese Society for Artificial Intelligence, 2019
Number : 33
Location : [in Japanese]
Date : June 04, 2019 - June 07, 2019
Black-box optimization is a problem of optimizing the objective function within the bounds of a given budget for evaluations. In Black-box optimization, it is generally assumed that the calculation cost for evaluating one solution is large, so it is important to search efficiently with as few budgets as possible. However, there is a problem that state-of-the-art black-box optimization methods such as Bayesian optimization and CMA-ES do not consider a budget. In this paper, we aim to propose an initialization of a search space that takes a budget into account, dealing with the above problem. We confirm that the proposed method shows good performance by experiments on the benchmark functions.