# Algorithm Configuration

Given a parameterized algorithm $\mathcal{A}$ with possible parameter settings $\mathcal{C}$, a set of training problem instances $\mathcal{I}$, and a performance metric $m: \mathcal{I} \times \mathcal{C} \rightarrow \mathbb{R}$, the algorithm configuration problem is to find a parameter configuration $c \in \mathcal{C}$ that minimizes $m$ across the instances in $\mathcal{I}$.