Algorithm Configuration Problem

Automatic configuration of algorithmic parameters is an area of active research [16, 27], going back to the foundational work [33], published in 1976. The algorithm configuration problem (ACP) focuses on parametrized algorithms, deployed to solve instances of a given decision or optimization problem. It concerns the identification of algorithmic parameters yielding the best performance when the algorithm is run on its input. Formally, we let A be a configurable target algorithm. Its input consists of an instance of the problem being solved and an array of parameters; we call the latter “algorithmic configuration”.