Document worth reading: “Automated Algorithm Selection: Survey and Perspectives”

It has prolonged been seen that for just about any computational draw back that has been intensely studied, completely totally different conditions are best solved using completely totally different algorithms. This is particularly pronounced for computationally exhausting points, the place typically, no single algorithm defines the cutting-edge; instead, there is a set of algorithms with complementary strengths. This effectivity complementarity may very well be exploited in diverse strategies, actually one among which relies on the considered selecting, from a set of given algorithms, for each draw back event to be solved the one anticipated to hold out best. The exercise of robotically deciding on an algorithm from a given set known as the per-instance algorithm alternative draw back and has been intensely studied over the earlier 15 years, leading to predominant enhancements throughout the cutting-edge in fixing a rising number of discrete combinatorial points, along with propositional satisfiability and AI planning. Per-instance algorithm alternative moreover reveals so much promise for enhancing effectivity in fixing regular and mixed discrete/regular optimisation points. This survey provides an abstract of study in automated algorithm alternative, ranging from early and seminal works to newest and promising utility areas. Different from earlier work, it covers capabilities to discrete and regular points, and discusses algorithm alternative in context with conceptually related approaches, harking back to algorithm configuration, scheduling or portfolio alternative. Since informative and cheaply computable draw back event choices current the premise for environment friendly per-instance algorithm alternative applications, we moreover current an abstract of such choices for discrete and regular points. Finally, we provide views on future work throughout the area and give attention to various open evaluation challenges. Automated Algorithm Selection: Survey and Perspectives