Meta-Optimization for Parameter Tuning with a Flexible Computing Budget

Meta-optimization techniques for tuning algorithm parameters usually try to find optimal parameter settings for a given computational budget allocated to the lower-level algorithm. If the available computational budget changes, parameters have to be optimized again from scratch, as they usually depend on the available time. For example, a small computational budget requires a focus on exploitation, while a larger budget allows more exploration. In situations where the optimization problem is expected to be solved for various computational budgets, meta-optimization is very time consuming. The method proposed in this paper can, in a single run, identify the best parameter settings for all possible computational budgets up to a specified maximum, hence saving a lot of time.