In the limit of low amplitude noise, the transition between metastable states is rare, and will most likely to follow an ``instanton'' trajectory. The probability of transition can be calculated from the instanton trajectory. However, for complicated systems it may be impractical to find the instanton. Alternatively, one can calculate an optimal set of $N$ discrete perturbations which lead to the basin boundary of the target state. I will show that this optimization problem is identical to calculating the instanton in the limit of $N\rightarrow\infty$, provided one uses the correct norm. We demonstrate this explicitly for the one-dimensional Swift--Hohenberg equation. Even when $N$ is small, the optimization problem gives an approximation to the more complicated instanton trajectory.