Automated Dynamic Algorithm Configuration

The performance of an algorithm often critically depends on its parameter configuration. While a variety automated configuration methods have been proposed to relieve users from the tedious and error-prone task manually tuning parameters, there is still lot untapped potential as learned static, i.e., settings remain fixed throughout run. However, it has shown that some parameters are best adjusted dynamically during execution. Thus far, this most commonly achieved through hand-crafted heuristics. A promising recent alternative automatically learn such dynamic adaptation policies data. In article, we give first comprehensive account new field (DAC), present series advances, provide solid foundation for future research in field. Specifically, (i) situate DAC broader historical context AI research; (ii) formalize computational problem; (iii) identify used prior art tackle (iv) conduct empirical case studies using evolutionary optimization, planning, machine learning.