local corrector #2
Description
Say you computed a statistic on a chunk, or only on a subset of digital samples, and you "detected" something (like a local max, or something above a threshold).
Now you may want to fine tune the location of this detection, searching before and after your initial crude estimate -- for example by wiggling around that location to find a better (possibly incomplete search, such as gradient search for a better local optimum), or a best (full search) score.
This technique falls under a more general framework of iterative improvements. In fact, peruse is meant for exactly that: making quick (and dirty) analysis of many chunks, then releasing them for further analysis down the pipeline.